You are on page 1of 48

MAIWALD · Elisenstraße 3 · 80335 München Maiwald Patentanwalts- und

Rechtsanwaltsgesellschaft mbH
The Comptroller
München
UK Intellectual Property Office (UKIPO) Düsseldorf
Concept House Elisenstraße 3
Cardiff Road 80335 München
Newport T +49 89 7472660
F +49 89 776424
South Wales H www.maiwald.eu
NP10 8QQ
UNITED KINGDOM

Munich, 29 December 2020 Dr. Simon Quartus Lud


Partner
lud@maiwald.eu
Official Ref.: 12 734 516.3 / 2 664 149
Patentee: Velos Media International Limited

Our Ref.: U07949NI/SQL

Grounds of Revocation Geschäftsführer


(Board of Directors)

Dr. Regina Neuefeind, LL.M.


Lutz Kietzmann, LL.M.
Dr. Norbert Hansen
Dr. Martin Huenges, LL.M.
This application for revocation is filed against validated European Patent EP 2 Dr. Holger Glas, LL.M.
Dr. Dirk Bühler
Dr. Christian Schäflein
664 149 B1 entitled “DEBLOCKING FILTERING” in the name and on behalf of Dr. Alexander Schmitz
Angela Zumstein
the Applicant for revocation (“the Applicant”), Dr. Derk Vos
Dr. Berthold Lux
Dr. Eva Ehlich
Alexander Ortlieb
Dr. Andreas Ledl
Unified Patents, LLC Dr. Eva Dörner
Dr.-Ing. Sophie Ertl
Dr. Marco Stief, LL.M.
Dr. Ralf Kotitschke
1875 Connecticut Ave., NW, Floor 10 Dr. Christian Pioch

HRB Nr. 111307


Washington, D.C. 20009 Amtsgericht München

SQL:vs
Table of contents

A. Introduction and Formalities ................................................................... 3


1. Requests..................................................................................................... 3
2. The Prior Art and further documents as provided ...................................... 4

B. European Patent (UK) EP 2 664 149 B1 ................................................. 6


1. Formalities .................................................................................................. 6
2. Subject matter of the alleged invention ...................................................... 6
2.1 Blocking Artefact v. Natural Edge ....................................................... 6
2.2 Alleged Problem .................................................................................. 8
2.3 Representative Claim ........................................................................ 12
2.4 Level of Ordinary Skill in the Art........................................................ 13

C. Grounds for Revocation ........................................................................ 13


I. Lack of novelty and inventive step over D3 ............................................ 13
1. Overview of cited document D3 ............................................................... 13
2. Lack of novelty and lack of inventive step of claim 1 with regard to cited
document D3.................................................................................................... 18
3. Lack of lack of inventive step of dependent claim 2 with regard to a
combination of cited document D3 and cited document D5............................ 25
4. Lack of lack of inventive step of dependent claim 3 with regard to a
combination of cited document D3 and cited document D5............................ 28
5. Lack of novelty and lack of inventive step of claim 7 with regard to cited
document D3.................................................................................................... 31
6. Lack of novelty and lack of inventive step of claim 22 with regard to cited
document D3.................................................................................................... 32
II. Lack of novelty and inventive step over D4 ............................................ 33
1. Overview of cited document D4 ............................................................... 33
2. Differences between cited document D4 and cited document D3 ........... 39
3. Lack of novelty and lack of inventive step of claim 1 with regard to cited
document D4.................................................................................................... 40
4. Motivation to “Modify…Two Consecutive Pixels” in the “Default Mode” .. 45

D. Summary and Conclusion ..................................................................... 48

2
A. Introduction and Formalities

1
The Applicant hereby requests revocation of validated European Patent
EP 2 664 149 B1 (hereinafter “the Patent”) granted with effect from 16
November 2016 (Section 25(1)) and validated to
VELOS MEDIA INTERNATIONAL LIMITED,
Unit 32, The Hyde Building
The Park
Carrickmines
Dublin 18
Ireland
[ADP Number 12041828001]

2
VELOS MEDIA INTERNATIONAL LIMITED is hereinafter referred to as
“the Patentee”.

1. Requests

3 It is hereby requested (1) that the Patent be revoked in its entirety under
section 72(1) of The Patents Act 1977 (“the 1977 Act”) on the grounds of:

(a) section 72(1)(a) because the subject matter of the Patent lacks
novelty contrary to sections 1(1)(a), 2 and/or lacks inventive step
contrary to sections 1(1)(b), 3.

4
In the event that the comptroller is not in a position to revoke the Patent
in its entirety (see request (1) above), a hearing is hereby requested.

3
2. The Prior Art and further documents as provided

5
In these the grounds of revocation, the applicant for revocation relies upon
the following documents:

Ref. Publication No.

D01 US Patent 9,414,066

D02 Declaration of Dr. Lina Karam

D03 US Patent No. 6,028,967

Sung Deuk Kim; Jaeyoun Yi; Hyun Mun Kim;


Jong Beom Ra, “Deblocking Filter with Two
D04 Separate Modes in Block-Based Video
Coding,” IEEE Trans. On Cir. and Sys. for
Video Tech., Feb. 1999

D05 U.S. Publication No. 2006/0008038

“H.264 and MPEG-4 Video Compression,” Iain


D06 E.G. Richardson, 2004, excerpts of chapters
1, 2, 3, 5, 6, and 7

D07 File History—Office Action, February 18, 2015

D08 File History—Amendment, May 18, 2015

D09 File History—Office Action, June 10, 2015

File History—After Final Response, July 23,


D10
2015

D11 File History—Advisory Action, July 30, 2015

4
File History—Pre-Appeal Brief Request for
D12
Review, September 8, 2015

File History—Notice of Panel Decision from


D13
Pre-Appeal Brief Review, November 4, 2015

D14 File History—Appeal Brief, December 3, 2015

File History—Notice of Allowance, April 1,


D15
2016

Li et al., “An Efficient Deblocking Filter


Architecture with 2-Dimensional Parallel
D16 Memory for H.264/AVC,” IEEE Proceedings of
the Asia and South Pacific Design Automation
Conference, 2005 (“Li”)

D17 Declaration of Dr. Sylvia Hall-Ellis

6
The list of references is annexed hereto as

Annex M1.

7 Each of the prior art documents D3 to D6 and D16 of all documents D1 to


D17 listed above was published before the earliest priority date of the
Patent, which is 14.01.2011. Therefore, each of cited documents D3 to
D6, and D16 constitutes prior art under Article 54(2) EPC.

8
Thus, the prior art documents D3 to D6 and D16 are all relevant for the
assessment of both lack of novelty and inventive step.

9
Since the vast majority of the cited prior art documents D3 to D6 and D16
was available to the public prior to the claimed priority date, the issue of
entitlement to priority is not discussed in detail herein below. However, this
is no admission that the by the Patent as granted after examination
procedure before the EPO claimed subject matter would validly claim the
priority.
10
References to the cited prior art documents D3 to D6 and D16 are to
paragraphs or sections of the publications listed above. These references,
however, are exemplary and not exhaustive. The the Applicant for
revocation reserves the right to refer to other sections of the prior art
documents if necessary later in the proceedings.

5
B. European Patent (UK) EP 2 664 149 B1

The legal status of the European Patent EP 2 664 149 B1 is discussed in


the following:

1. Formalities

11
The European Patent (UK), EP 2 664 149 B1, derives from European
Patent Application Number EP12734516.3 (hereinafter “the European
Application”) which has a filing date of 13th of January 2012.

12 The European Application was published as EP 2 664 149 A1 on


20.11.2013. The European Application was published as A1 Application
with search report.

13
The Patent claims the priority dates of 14.01.2011 and 02.03.2011, which
are the filing dates of US provisional Patent Applications
US201161432746P and US201161448296P (hereinafter “the Priority
Applications”).

14
The European Application is the regional family member of the
international patent application WO2012096623.

15
In the following, references to the Patent and paragraphs thereof – shown
in square brackets “[…]” – relate to the published B1-specification, unless
specified otherwise.

2. Subject matter of the alleged invention

16
The alleged invention of the Patent is generally directed to deblocking
filtering techniques providing pixel values of pixels located in close vicinity
of a block boundary to selectively be modified or maintained unmodified
depending on the relative difference between the considered pixel values.
See the Patent, [0005]-[0006].

2.1 Blocking Artefact v. Natural Edge

17
As described in the Patent regarding video encoding and decoding,
deblocking filtering is preferably applied “only…over block boundaries
where there are blocking artifacts.” See the Patent, [0090]. Deblocking
filtering should not be applied (or the amount of filtering should be
reduced) “in situations where the block boundary between two

6
neighbouring [sic] blocks is considered to be a natural edge”. See the
Patent, [0028].

18
The difference between a blocking artefact and a naturally-occurring edge
is illustrated in Figures 1a and 1b (below) of the Patent using two
examples of pixel values across a block boundary, vertical grey line in
Figure 1a and Figure 1b of the Patent as reprinted below. The horizontal
axis represents pixel position and the vertical axis represents the pixel
value magnitude.

Figure 1a of the Patent

Figure 1b of the Patent

7
19
Figure 1a of the Patent illustrates pixel values 10a, 11a, and 12a of a first
block and pixel values 13a, 14a, and 15a of a neighbouring block. See the
Patent, [0070].

20
Similarly, Figure 1b illustrates pixel values 10b, 11b, and 12b of a first
block and pixel values 13b, 14b, and 15b of a neighbouring block. See the
Patent, [0071]. In both figures, the first block and the neighbouring block
are separated by a block boundary.

21
The Patent proposes that the large difference between the pixel values on
either side of the block boundary in Figure 1a is of a natural origin rather
than a blocking artefact. See the Patent, [0070].

22
For example, this natural origin can be due to the presence of a bright
object against a dark background, which creates a large and abrupt
transition in the pixel values at the object’s edges. Such a natural edge
“should only be exposed to a limited filtering or no filtering at all”. See the
Patent, [0069].

23
However, the relatively smaller difference in pixel values on either side of
the block boundary in Figure 1b likely results from quantization and/or
prediction used in video coding and is thus a blocking artefact that
requires filtering. See the Patent, [0071].

24
The background section of the Patent summarizes two known deblocking
filtering methods, one corresponding to the H.264 standard, see the
Patent, [0003], and another to the HEVC standard, see the Patent, [0013].

25
Contrasting the two methods, the Patent contends that “[i]n H.264 filtering,
there is a check…for the purpose [of] deciding whether there is a blocking
artefact [sic] at the block boundary or if it is a natural edge at the block
boundary.”. See the Patent, [0022].

26
Conversely, the HEVC’s proposed deblocking filtering method does not
perform such a check, and thus undesirably filters natural edges. See the
Patent, [0022].

2.2 Alleged Problem

27
The Patent contends that the reason that HEVC’s filtering method does
not check for natural edges in the same manner as H.264 is because, in
HEVC, “deblocking filtering is done also [when] the signals [resemble]
inclined ramps”. See the Patent, [0022].

8
28
The Patent describes a “ramp” as “linearly increasing or decreasing pixel
values [across a block boundary], e.g. 10, 20, 30, 40”. See the Patent,
[0105].

29
According to the technical disclosure of the Patent, a “ramp” is contrasted
with a “step,” which the Patent describes as “a step increase or decrease
in pixel values [across a block boundary], e.g. 10, 10, 20, 20”. See the
Patent, [0105]. For illustrative purposes, these examples are charted
below.

30
In HEVC, both ramp-like and step-like transitions across block boundaries
are filtered normally. See the Patent, [0022]. This is a problem, according
to the Patent, because step-like transitions with large step sizes likely
correspond to natural edges and should not be filtered (or should be
subject to reduced filtering). See the Patent, [0071].

31
The Patent therefore proposes that HEVC should include a check for
natural edges (i.e., a check that prevents filtering of large, step-like
transitions), as is done in H.264. See the Patent, [0022]. However, the
H.264 check for natural edges uses only the values of pixels positioned
directly along the block boundary. See the Patent, [0071].

32
Such a check cannot distinguish a “step” of a natural edge from a “ramp”
that should be filtered normally. See the Patent, [0009]. For example, in
the above examples of a “ramp” and a “step,” the difference between pixel
values of the pixels directly on either side of the block boundary is the
same in both cases. Therefore, this difference provides no basis for
distinguishing a “ramp” from a “step.”

9
33
Consequently, the check performed in H.264 is not suitable for use in
HEVC. See, e.g., the Patent, [0022].

34
The Patent suggests solving this drawback by checking for natural edges
across block boundaries using two pixels from each of two adjoining
blocks, for a total of four consecutive pixels.

35
Unlike the H.264 check, which uses only one pixel from each adjoining
block, the Patent check uses enough pixels to distinguish between a
“ramp” and a “step” transition across the block boundary. See the Patent,
[0105].

For example, Figure 2a of the Patent, reproduced below, depicts a line of


pixels surrounded with a grey rectangular frame structure, the line set in
relation to vertical block boundary 20. See the Patent, [0074].

Figure 2a of the Patent.

36
The surrounded line of pixels contains two consecutive pixels from a first
block 21, p0 and p1 (subscripts omitted), and two consecutive pixels from
a neighboring block 22, q0 and q1 (subscripts omitted).

37
First block 21 and neighboring block 22 are located on opposite sides of
block boundary 20 and are next to block boundary 20. Pixels p1, p0, q0
and q1 form a line of pixels that is perpendicular to the block boundary 20.
Similar observations may be made for a horizontal block boundary. See
the Patent, Figure 2b.

10
38
In one embodiment, the method checks for natural edges by calculating
an offset—“first offset”—based on four consecutive pixel values as
follows:

39
first offset=((9×(q0-p0)-3×(q1-p1)))⁄16 (1)

40
See the Patent, [0085]. p1, p0, q0, and q1 denote pixel values for four
consecutive pixels. See the Patent, [0082] and Figs. 2a-2b of the Patent.

41
The first term, “9×(q0-p0)” amplifies the difference between the two pixels
closest to the block boundary on either side of the block boundary – that
is, the step size directly across the block boundary.

42
The second term, “3×(q1-p1)” amplifies (by a smaller amount) and
subtracts from the first term the step size between the pixels one position
further away from the block boundary.

43
For a step-like transition, the step size is about the same in both the first
term and the second term, so the first term predominates over the second
term due to the larger multiplier – i.e., 9 versus 3.

44
This results in a relatively large value of the first offset. On the other hand,
for a ramp-like transition, the step size of the second term is larger than
the step size of the first term because the step size increases when
moving away from the block boundary.

45
As a result, the second term counteracts the first term despite its smaller
multiplier, yielding a smaller first offset value.

46 Consequently, the output of equation (1) distinguishes between a step-


like transition (i.e., offset) across the block boundary and other types of
transitions, such as ramp-like transitions. See the Patent, [0085].

47 To illustrate, when p1=10, p0=10, q0=20, and q1=20 (i.e., step-like


transition), the “first offset” is 3.75. See the Patent, [0105]. Meanwhile,
when p1=10, p0=20, q0=30, and q1=40 (i.e., ramp-like transition), the
“first offset” is zero. See the Patent, [0085]. Similarly, the “first offset” is
zero if the pixel values are equal (p0=p1=q0=q1), forming a straight line.
See the Patent [0089].

Next, the method compares the “first offset” to a threshold (thr1) to


determine whether to apply normal or reduced filtering, Figures 3a-3c. For
example, as shown below, when the absolute value of the “first offset” is
smaller than the threshold, a set of filtering operations is applied to the
pixel values. See the Patent, [0085]. Each of p1, p0, q0, and q1 is

11
modified based on the “first offset” to yield filtered values p1’, p0’, q0’, and
q1’.

Conversely, when the “first offset” is larger than the threshold, the ’066
Patent proposes applying reduced filtering or no filtering to the pixel
values.

2.3 Representative Claim

48 Claim 1 is representative and recites, among other limitations, “calculating


(S2) a first offset based on the pixel values of the four pixels (41,42,45,46),
constituting two pixel pairs, each pixel pair comprising corresponding
pixels from both blocks, by determining the relative difference between
the pixel values of the first pixel pair (41,45) and the relative difference
between the pixel values of the second pixel pair (42,46);” and “modifying
(S4) the respective pixel values of the first pixel pair (41,45), by adding
said first offset to one of said consecutive pixels (41) and by subtracting
said first offset from the other of said consecutive pixels (45), in case
abs[first offset] < first threshold, or not modifying any pixel at all, in case
abs[first offset] >= first threshold.”

49
Independent claim 7 of the Patent, relating to a deblocking filter unit for
reducing blocking artefacts, recites similar elements compared to the
method steps of claim 1 of the Patent.

50
Although claim 1 of the Patent recites “calculating a first offset,” it does
not identify the equation used to calculate the “first offset.”

51
By contrast, dependent claim 6 of the Patent recites equation (1) for
calculating the “first offset”. Accordingly, the “calculating a first offset”
limitation of claim 1 is broader than using equation (1).

12
As shown in the following, all the features or limitations of claim 1 of the
Patent were known in the art, and were likewise obvious before the
Patent’s priority date.

As explained in detail below, cited document D3 and cited document D4


each disclose calculating an offset as a function of at least two
consecutive pixel values from a first block and at least two consecutive
pixel values from a neighboring block, comparing the offset to a threshold,
and modifying pixel values by applying filtering, if the absolute value of
the offset is less than the threshold. See cited document D3, 4:6-65; see
cited document D4, Fig. 2.

Moreover, as the Patent admits, calculating such an offset is known, see


background section of the Patent.

Cited document D3 and cited document D4 each further disclosed


modifying at least two consecutive pixel values from the first block and at
least two consecutive pixel values from the neighboring block. See cited
document D3, 4:50-65; see cited document D4, 158.

2.4 Level of Ordinary Skill in the Art

A person of ordinary skill in the art at the time of filing the earliest
provisional application for the Patent would have had at least the
equivalent of a Bachelor’s degree in electrical engineering, computer
engineering, or a related subject and two or more years of experience in
the field of video coding. Less work experience may be compensated by
a higher level of education, such as a Master’s Degree.

C. Grounds for Revocation

I. Lack of novelty and inventive step over D3

52
The subject matter of each of the independent claims 1 and 7 of the Patent
is disclosed by or at least rendered to be non-obvious over cited document
D3 for the following reasons.

1. Overview of cited document D3

53
Cited document D3 is directed to a “method of reducing a blocking artifact
when coding moving picture.” See cited document D3, Title, 2:28-30. See

13
also cited document D3, 1:60-64 (“In processing a moving picture...blocks
are individually processed”).

54
Further, also the problems to be solved of the Patent and cited document
D3 match: See cited document D3, 2:27-30 “An object of the present
invention is to provide a method for reducing a blocking artifact“ on the
one hand and one the other hand, in corresponding manner, the Patent
defines “It is therefore a need for an efficient blocking filter that can be
used to reduce blocking artifacts at block boundaries, but which does not
have the above mentioned drawbacks.”. See the Patent at [0023].

55
Accordingly, cited document D3 is in the same field of invention as the
Patent (video coding) and is directed to the same problem to be solved
(reducing blocking artefacts).

56
Cited document D3 is therefore a prior art document in the same or at
least a neighbouring technical field of the Patent.

57
Figure 1 of cited document D3, reproduced below, depicts a line of pixels
S0 in relation to horizontal and vertical block boundaries. See cited
document D3, 3:11-21.

58
According to the technical teaching of cited document D3, the line of pixels
S0 “is located across a block boundary” and thus “is directly influenced by
the blocking artifact.” See cited document D3, 3:17-19.

59 The line of pixels S0 contains two consecutive pixels from a first block, v2
and v3, and two consecutive pixels from a neighbouring block, v4 and v5.
The first block and the neighbouring block are located on opposite sides
of the block boundary and are next to the block boundary.

60 The pixels v2, v3, v4, and v5 form a line of pixels that is perpendicular to
the block boundary.

61
A comparison below between the Patent’s Figures 2a and 2b and Figure
1 cited document D3, each annotated similarly, shows that pixel values
v2, v3, v4, and v5 surrounded with a grey rectangular frame structure of
cited document D3 correspond to pixel values p1, p0, q0, and q1 of the
Patent, which are also surrounded with a grey rectangular frame structure
in the picture given below.

14
The Patent, Fig. 2a

The Patent, Fig. 2b Cited document D3, FIG. 1

62
Cited document D3 suggests using a frequency analysis tool, such as one
based on the discrete cosine transform, DCT, to analyse and reduce
blocking artefacts. See cited document D3, 3:46-48.

63
Figure 2 of cited document D3, below, illustrates the DCT basis functions
used to calculate corresponding DCT coefficients for the line of pixels S0
identified in Figure 1. See cited document D3, 3:49-53.

15
Cited document D3, FIG. 2

64
The DCT basis functions each include four points (represented by stem
points) and vary according to frequency and symmetry with respect to the
block boundary. See cited document D3, 3:49-52.

65
The technical teaching of cited document D3 notes that the fourth DCT
coefficient a3,0 “directly affect[s] the block discontinuity.” See cited
document D3, 3:18-20 (“The blocking artifact appears…in the form of a
line of discontinuity”).

66 Analogously, the Patent explains that the “first offset” is related to the
“relative difference” between pixel values on either side of a block
boundary. See the Patent, [0028].

67 Accordingly, both DCT coefficient a3,0 of the cited document D3 and the
Patent’s “first offset” describe a “discontinuity” or “difference” in pixel
values across the block boundary.

68 Cited document D3 discloses that a3,0 is “evaluated from the simple inner
product of the DCT kernel and pixels,” as follows:

𝑎3,0 = ([𝑐1 − 𝑐2 𝑐2 − 𝑐1 ] ∗ [𝑣2 𝑣3 𝑣4 𝑣5 ]𝑇 )⁄𝑐3 (2)

69
See cited document D3, 4:28-32. Equation (2) of Cited document D3 may
be expressed equivalently1 as:

1 Evaluating the inner product in the numerator of equation (2) yields the following
intermediate equation:
𝑎3,0 = (𝑐1 × 𝑣2 − 𝑐2 × 𝑣3 + 𝑐2 × 𝑣4 − 𝑐1 × 𝑣5 )⁄𝑐3

16
𝑎3,0 = (𝒄𝟐 × (𝑣4 − 𝑣3 ) − 𝒄𝟏 × (𝑣5 − 𝑣2 ))⁄𝒄𝟑 (3)

70 As explained above, v2, v3, v4, and v5 denote pixel values for two
consecutive pixels from the first block (v2 and v3) next to the block
boundary and two consecutive pixels of the neighbouring block (v4 and v5)
and correspond to p1, p0, q0, and q1 of the Patent.

71
Cited document D3 explains that c1 and c2 are preferably approximated
by integers and c3 is preferably approximated to a multiple of two.
EX1003, 4:24-27. ’066 Patent’s equation (1) is provided for comparison:

first offset = (𝟗 × (𝑞0 − 𝑝0) − 𝟑 × (𝑞1 − 𝑝1))⁄𝟏𝟔 (1)

72
Comparing the Patent’s equation (1) to equations (2) and (3) of cited
document D3, equation (1) assigns numerical values 3, 9, and 16 to
constants c1, c2, and c3 and is otherwise identical.

73 Consistent with cited document D3, c1 and c2 are integers (3 and 9) and
c3 is a multiple of 2 (i.e., 16). Thus, DCT coefficient a3,0, of cited document
D3 calculated from equation (2), corresponds to the Patent’s first offset as
calculated from equation (1).

74
The technical disclosure of cited document D3 teaches modifying the pixel
values based on a comparison of a threshold to the DCT coefficient a3,0.
In particular, cited document D3 discloses modifying pixel values v2, v3,
v4, and v5, as follows:

𝑣3′ = 𝑣3 − 𝑑[1] (4)2

𝑣4′ = 𝑣4 + 𝑑[1] (5)

𝑣2′ = 𝑣2 − 𝑑2 (6)

v5′ = v5 + d2 (7)

See cited document D3. In equations (4)–(7):

𝑑1 = (3 (𝑣3 − 𝑣4 )⁄8) × (𝛿(|𝑎3,0 | < QP)) (8)

𝑑2 = (3 (𝑣3 − 𝑣4 )⁄16) × (𝛿(|𝑎3,0 | < QP)) (9)

1
Grouping the numerator terms by constants 𝑐1 and 𝑐2 yields equation (3).
2Omission of the subscript from d1 in this equation of cited document D3
appears unintentional.

17
75 See cited document D3, 4:60-65. The function 𝛿(|𝑎3,0 | < QP) compares
the absolute value of a3,0 (first offset) to QP, i.e., “the quantization
parameter of the macroblock where v4 belongs.” See cited document D3,
4:21-23. When the absolute value of a3,0 (first offset) is less than QP, then
𝛿(|𝑎3,0 | < QP) = 1. Otherwise, 𝛿(|𝑎3,0 | < QP) = 0.

76
Thus, quantization parameter QP of cited document D3 operates as a
threshold (first threshold). See the Patent, [0100] (“the first threshold is
determined on the basis of quantization parameter values assigned to
the respective block”).

77
When the absolute value of a3,0 (first offset) is less than QP (first
threshold), then 𝛿(|𝑎3,0| < QP) = 1, and d1 and d2 are nonzero.
Consequently, equations (4)-(7) apply filtering to the block boundary such
that “[t]he blocking artifact…is reduced.” See cited document D3, 4:66-
67.

78 That is, the values of the pixels of the first block are modified by an amount
-d1 or -d2 (e.g., the pixel values are decreased if d1/d2 are positive) and
the values of the pixels of the neighbouring block are modified by a
corresponding amount with opposite sign, +d1 or +d2 (e.g., the pixel values
are increased if d1/d2 are positive).

79
This reduces the difference between the pixel values on opposite sides of
the block boundary (i.e., their modified values are closer together than the
unmodified values) and thus filters the blocking artifact.

80
On the other hand, when the absolute value of a3,0 is greater than QP,
(|𝑎3,0 | ≥ QP is true), then 𝛿(|𝑎3,0 | < QP) = 0, and d1 and d2 are zero.
Equations (4)–(7) yield v2'=v2, v3'=v3, v4'=v4, and v5'=v5. Consequently,
equations (4)-(7) apply no filtering on pixels v2, v3, v4, and v5.

2. Lack of novelty and lack of inventive step of claim 1 with regard to


cited document D3

81 Claim 1 of the Patent recites as follows in the beginning:

“A method of reducing blocking artifacts associated with consecutive


pixels of a block boundary (20) of an image, the method comprising”

82
To the extent that the indication of purpose is deemed limiting, cited
document D3 discloses, or renders obvious, the above recited feature.

18
83
Specifically, cited document D3 discloses a “method of reducing a
blocking artifact when coding moving picture,” i.e., a sequence of digital
images. See cited document D3, Title, 2:28-30.

84
For a claim that is directed to a method or process, the indication of an
intended use of this method may at most be seen as limiting to the extent
that the method has to be suitable for that use.

85
Such a claim would therefore be anticipated by a prior-art document
describing a method having such suitability although not mentioning the
specific use

86
Cited document D3 further discloses “an MPEG-4 video coding method
that reduces a blocking artifact in a real-time moving picture using a
frequency property around boundaries between blocks.” See cited
document D3, 2:32-35. See also cited document D3, 4:38-39, 5:40-41.

87
In particular Figure 1 of cited document D3, reproduced below, depicts a
line of pixels S0 surrounded with a grey rectangular frame structure in
relation to horizontal and vertical block boundaries of a digital image. See
cited document D3, 3:11-15.

88
In the technical disclosure of cited document D3 the line of pixels S0 “is
located across a block boundary” and thus “is directly influenced by the
blocking artifact.” See cited document D3, 3:17-19.

89 The line of pixels S0 contains two consecutive pixels from a first block,
v2 and v3, and two consecutive pixels from a neighbouring block, v4 and
v5.

90
A comparison below between the Patent’s FIG. 2a and FIG. 1 of cited
document D3, each annotated similarly, shows that pixel values v2, v3, v4,
and v5 of cited document D3 correspond to pixel values p1, p0, q0, and
q1 of the Patent, which are also surrounded with a grey rectangular frame
structure.

19
The Patent, Fig. 2a

The Patent, Fig. 2b Cited document D3, FIG. 1

91 Similar observations may be made for the horizontal block boundary in


Figure 1 of cited document D3 and the Patent’s Figure 2b, reproduced
above with similar annotations.

92
Thus, cited document D3 discloses, or renders obvious, the above recited
features of claim 1 of the Patent.

93
Claim 1 of the Patent further claims the following feature:

“selecting (S1) at least two consecutive pixels (41,42) from a first


block and at least two consecutive pixels (45,46) from a
neighboring block, the blocks being located on opposite sides of a
block boundary (20) the consecutive pixels (41,42,45,46) being
located next to the block boundary, and the pixels forming a line of
pixels (41,42, 45,46) which is perpendicular to the block boundary
(20);”

94
Cited document D3 discloses, or renders obvious, this limitation. Figure 1
of cited document D3, reproduced below, depicts a line of pixels S0
positioned relative to a horizontal and vertical block boundaries. See cited
document D3, 3:11-15. The technical disclosure of cited document D3
notes that the line of pixels S0 “is located across a block boundary” and

20
thus “is directly influenced by the blocking artifact.” See cited document
D3, 3:17-19.

95
The line of pixels S0 contains two consecutive pixels from a first block,
v2 and v3, and two consecutive pixels from a neighbouring block, v4 and
v5. The first block and the neighbouring block are located on opposite
sides of the block boundary and next to the block boundary.

96
The pixels v2, v3, v4, and v5 form a line of pixels that is perpendicular to
the block boundary. A comparison below between the Patent’s Figures
2a and 2b and Figure 1 of cited document D3, each annotated similarly,
shows that pixel values v2, v3, v4, and v5 of cited document D3 correspond
to pixel values p1, p0, q0, and q1 of the Patent.

The Patent, Fig. 2a

The Patent, Fig. 2b Cited document D3, FIG. 1

97 Cited document D3 therefore discloses, or at least renders obvious, the


subsequent limitation in terms of the following feature of claim 1 of the
Patent:

“calculating (S2) a first offset based on the pixel values of the four
pixels (41,42,45,46), constituting two pixel pairs, each pixel pair
comprising corresponding pixels from both blocks, by determining
the relative difference between the pixel values of the first pixel

21
pair (41,45) and the relative difference between the pixel values of
the second pixel pair (42,46);”

98
Cited document D3 discloses, or renders obvious, this limitation. As
discussed above, the technical teaching of cited document D3 discloses
that the DCT coefficient a3,0, corresponding to the line of pixels S0, affects
“block discontinuity.” See cited document D3, 3:57-623.

99
Analogously, the Patent explains that the “first offset” is related to the
“relative difference” between pixel values on either side of a block
boundary. See claim 1 feature recited above and see the Patent, [0028].

100 Accordingly, both DCT coefficient a3,0 as described in cited document D3


and the Patent’s “first offset” describe a “discontinuity” or “difference” in
pixel values across the block boundary.

101
Cited document D3 further discloses calculating the DCT coefficient a3,0
(first offset) as a function of pixel values v2, v3, v4, and v5 as follows:

𝑎3,0 = ([𝑐1 − 𝑐2 𝑐2 − 𝑐1 ] ∗ [𝑣2 𝑣3 𝑣4 𝑣5 ]𝑇 )⁄𝑐3 (2)

102
Cited document D3, 4:28-32. Equation (2) of cited document D3 may be
expressed equivalently4 as:

𝑎3,0 = (𝑐2 × (𝑣4 − 𝑣3 ) − 𝑐1 × (𝑣5 − 𝑣2 ))⁄𝑐3 (3)

103 As explained above, v2, v3, v4, and v5 denote pixel values for two
consecutive pixels from the first block (v2 and v3) and two consecutive
pixels of the neighbouring block (v4 and v5) and correspond to pixel values
p1, p0, q0, and q1 of equation (1). Cited document D3 explains that c1
and c2 are preferably approximated by integers and c3 is preferably
approximated to a multiple of two. See cited document D3, 4:24-27.

104 Although claim 1 does not require a particular equation to calculate the
“first offset,” equation (1) of the Patent is replicated here for comparison:

105 first offset = (9 × (𝑞0 − 𝑝0) − 3 × (𝑞1 − 𝑝1))⁄16 (1)

106
Comparing equation (1) of the Patent to equations (2) and (3) of cited
document D3, equation (1) assigns numerical values 3, 9, and 16 to
constants c1, c2, and c3 and is otherwise identical. Consistent with cited
document D3, c1 and c2 are integers (i.e., 3 and 9) and c3 is a multiple of
2 (i.e., 16).

3 All emphasis added, unless noted otherwise.


4
See above for a step-by-step derivation.

22
107
Thus, a3,0 corresponds to the claimed first offset. Moreover, as shown in
equations (2) and (3), a3,0 is calculated as a function of pixel values of two
consecutive pixels from the first block and two consecutive pixels of the
neighbouring block.

108
Cited document D3 therefore discloses, or at least renders obvious, the
subsequent limitation in terms of the following feature of claim 1 of the
Patent:

“comparing (S3) the first offset to a first threshold value, and”

109
Cited document D3 discloses, or renders obvious, this limitation. For
example, cited document D3 discloses comparing the DCT coefficient a3,0
(first offset) to a quantization parameter QP (first threshold) using the
following equations:

𝑑1 = (3 (𝑣3 − 𝑣4 )⁄8) × (𝛿(|𝑎3,0 | < QP)) (8)

𝑑2 = (3 (𝑣3 − 𝑣4 )⁄16) × (𝛿(|𝑎3,0 | < QP)) (9)

110
See cited document D3, 4:60-65. The function 𝛿(|𝑎3,0 | < QP) compares
the absolute value of a3,0 (first offset) to QP, i.e., “the quantization
parameter of the macroblock where v4 belongs.” See cited document D3,
4:21-23. When the absolute value of a3,0 (first offset) is less than QP, then
𝛿(|𝑎3,0 | < QP) = 1. Otherwise, 𝛿(|𝑎3,0 | < QP) = 0.

111
Thus, the quantization parameter QP of cited document D3 operates as a
threshold (first threshold). See also the Patent, [01000] (“the first
threshold is determined on the basis of quantization parameter values
assigned to the respective block”).

112 Cited document D3 further discloses comparing the DCT coefficient a3,0
(first offset) to the quantization parameter QP (first threshold).

113
Cited document D3 therefore discloses, or at least renders obvious, the
subsequent limitation in terms of the following feature of claim 1 of the
Patent:

“modifying (S4) the respective pixel values of the first pixel pair
(41,45), by adding said first offset to one of said consecutive pixels
(41) and by subtracting said first offset from the other of said
consecutive pixels (45), in case abs[first offset] < first threshold, or
not modifying any pixel at all, in case abs[first offset] >= first
threshold.”

23
114
Cited document D3 discloses, or renders obvious, this limitation. Cited
document D3 discloses modifying pixel values of two consecutive pixels
from the first block (v2 and v3) and pixel values of two consecutive pixels
of the second block (v4 and v5), as follows:

𝑣3′ = 𝑣3 − 𝑑[1] (4)

𝑣4′ = 𝑣4 + 𝑑[1] (5)

𝑣2′ = 𝑣2 − 𝑑2 (6)

𝑣5′ = 𝑣5 + 𝑑2 (7)

See cited document D3, 4:50-59, where:

𝑑1 = (3 (𝑣3 − 𝑣4 )⁄8) × (𝛿(|𝑎3,0 | < QP)) (8)

𝑑2 = (3 (𝑣3 − 𝑣4 )⁄16) × (𝛿(|𝑎3,0 | < QP)) (9)

115
See cited document D3, 4:60-65.

116
When |𝑎3,0 | < QP is true, then 𝛿(|𝑎3,0 | < QP) = 1, and d1 and d2 are
nonzero. Consequently, equations (4)-(7) apply filtering to the pixels v2,
v3, v4, and v5 such that “[t]he blocking artifact…is reduced.” See cited
document D3, 4:66-67.

117
That is, the values of the pixels of the first block are modified by an amount
-d1 or -d2 (e.g., the pixel values are decreased if d1/d2 are positive) and
the values of the pixels of the neighbouring block are modified by a
corresponding amount with the opposite sign, +d1 or +d2 (e.g., the pixel
values are increased if d1/d2 are positive).

118 This reduces the difference between the pixel values on opposite sides of
the block boundary (i.e., their modified values are closer together than the
unmodified values) and thus filters the blocking artifact.

119
On the other hand, when |𝑎3,0 | ≥ QP is true, then 𝛿(|𝑎3,0 | < QP) = 0, and
d1 and d2 are zero. Equations (4)–(7) result into v2'=v2, v3'=v3, v4'=v4, and
v5'=v5.

120
According to the Patent, a filtering operation changes the value of a pixel.
See, e.g., the Patent, [0006] (“filtering can be described with a delta value
with which filtering changes the current value”); see, e.g., the Patent,
[0079] (“filtering is executed by modifying the respective pixel values of
consecutive pixels”); see, e.g., the Patent, [0109] (“apply normal filtering

24
by modifying one or more pixel values”); see, e.g., the Patent, [0128]
(“filtering unit 81 is configured to filter the image by modifying the
respective pixels”).

121
Consequently, equations (4)-(7) of cited document D3, which suggest that
no pixel values have been modified, apply no filtering at all on any of pixels
v2, v3, v4, and v5 because the pixel values are not changed when |𝑎3,0 | ≥
QP.

122 To the extent Patent Owner argues that cited document D3 does not
explicitly disclose “applying no filtering at all” in the event that |𝑎3,0 | ≥ QP
is true, it would have been obvious to a person skilled in the art to apply
no filtering at all when |𝑎3,0 | ≥ QP.

123 As discussed above, when |𝑎3,0 | ≥ QP, the values of d1 and d2 are zero,
and the pixel values are not changed. Even if, for the sake of argument
such an operation is regarded as “filtering”, a person skilled in the art
would readily have understood that an operation that does not change the
input values is equivalent to and could alternatively be implemented by
“applying no filtering at all”.

124
Accordingly, it would have been obvious to modify method as disclosed in
cited document D3 to apply no filtering at all.

125 Thus, this limitation would have been obvious to a person skilled in the
art.

126
Therefore, cited document D3 discloses, or renders obvious, this
limitation.

127
Because cited document D3 discloses all the limitations of claim 1, cited
document D3 anticipates claim 1.

128
Alternatively, the differences between claim 1 and cited document D3, if
any, would have been obvious to a person skilled in the art and therefore
cited document D3 renders claim 1 obvious.

129
Therefore, the Patent is to be revoked in its entirety under section 72(1)(d)
of the 1977 Act.

3. Lack of lack of inventive step of dependent claim 2 with regard to a


combination of cited document D3 and cited document D5

130
Claim 2 of the Patent recites as follows:

25
“modifying (S40, S50) the respective pixel values of at least the two
pixels (41,45) located next to the block boundary (20) on the basis
of the first offset, and wherein, in case abs[first offset] < first
threshold value, the respective pixel values of the two pixels (42,46)
located one pixel away from the block boundary (20) are modified
(S41) on the basis of a second offset and a third offset, respectively.”

131 With regard to such intra prediction and using multiple thresholds in a
cascading way for modifying the pixels, cited document D5 discloses, or
at least renders obvious, that the pixel modification can be performed for
horizontal or vertical lines of neighbouring pixels in the block using
multiple thresholds for amending or updating the neighbouring pixels with
regard of whether the pixel are the last pixels before the block boundary,
the penultimate or the penultimate before the block boundary.

132 Cited document D5 discloses, or at least renders obvious, the following


feature of claim 2 of the Patent:

“modifying (S40, S50) the respective pixel values of at least the two
pixels (41,45) located next to the block boundary (20) on the basis
of the first offset”

133
This feature of claim 2 of the Patent is specifically disclosed by cited
document D5 by paragraph [0223]:

“[0223] In the H.264/AVC standard, intra prediction techniques


address cases where MCP fails. In intra prediction, the predictor
for a current block is obtained by spatial prediction from
neighboring reconstructed pixels. In 3D sub-band coding with an
open-loop MCTF structure, however, such spatial prediction tends
to introduce mismatch and error drifting. “

134
Thus, the above recited features of claim 2 of the Patent are described in
cited document D5, [0223].

135
Cited document D3 discloses, or at least renders obvious, the following
feature of claim 2 of the Patent:

“wherein, in case abs[first offset] < first threshold value, the


respective pixel values of the two pixels (42,46) located one pixel
away from the block boundary (20) are modified (S41) on the basis
of a second offset and a third offset, respectively”

136
The above recited feature of claim 2 of the Patent defines that further
thresholds are used for certain pixel modification of the non-boundary
pixels.

26
137
This is for instance disclosed in cited document D3. Claim 10 of cited
document D3 recites as follows:

“The method according to claim 6, wherein the adjusting step in the


second mode satisfies at least one of the following conditions:

v3'=v3 -d;

v4'=v4 +d;

v2'=v2 -d2;

v5'=v5 +d2;

v1'=v1 -d3 ; and

v6'=v6 +d3,

where

d1 =(3(v3 -v4)//8)*δ(|a3,0 |<QP),

d2 =(3(v3 -v4)//16)*δ(|a3,0 |<QP), and

d3 =(3(v3 -v4)//32)*δ(|a3,0 |<QP),

wherein v0 -v7 are initial boundary pixel values, v1 '-v6 ' are
adjusted boundary pixel values, a3,0 is the discontinuous
component of the discrete cosine transform coefficients of the first
pixel and QP is a quantization parameter of a macroblock
containing v4.

138
In summary, claim 10 of cited document D3 clearly teaches using 3
thresholds d, d2, and d3, grouping the pixel in pixels located next to the
block boundary and pixel located one pixel away from the block boundary,
and finally applying different thresholds d, d2, and d3 to these pixels
groups.

139
Further, in the alternative, cited document D5 also teaches using multiple
thresholds, as for instance discussed on paragraph [0022] of cited
document D5:

“0022. According to a fourth set of tools and techniques, a decoder


Such as a 3D Sub-band Video decoder computes one or more
update thresholds”

27
140
Claim 2 of the Patent clearly defines using multiple thresholds (in the
particular claimed embodiment three thresholds) to be used as a criteria
for modifying or updating a pixel. Using one or more update thresholds is,
however, disclosed in the above recited paragraph of cited document D5.

141
Thus, the additional features of claim 2 are described in cited document
D5, [0223].

142
Further, the in claim 2 of the Patent claimed condition – “abs[first offset]
< first threshold value” – can also be found in cited document D3, see in
particular 4:1-10 of cited document D3, which recites as follows:

“A method for reducing a blocking artifact in a default mode is as


follows:

v3 '=v3 -d;

v4 '=v4 +d; and

d=CLIP(c2 (a3,0 '-a3,0)//c3,0, (v3 -v4)/2)*δ(|a3,0 |<QP).

In the method, a3,0 '=SIGN(a3,0)*MIN(|a3,0 |,|a3,1 |,|a3,2 |), and q is the


component of DCT kernel. The condition |a3,0 |<QP is used to
count the influence of the quantization parameter on the blocking
artifact. The |a3,0| <QP condition also prevents over-smoothing
when the blocking artifact is not very serious. The clipping
operation on the compensated value prevents the direction of the
gradient at the boundary from being large or changed in an
opposite direction.”

143
Thus, it has been shown that the subject matter of claim 2 of the Patent is
rendered obvious by a combination of cited document D3 and D5.

144 The person skilled in the art would also have combined cited documents
D3 and D5 because both documents address the reduction of artefacts in
video coding and decoding.

145
Therefore, the subject matter of claim 2 is rendered obvious by a
combination of cited documents D3 and D5.

4. Lack of lack of inventive step of dependent claim 3 with regard to a


combination of cited document D3 and cited document D5

146
Claim 3 of the Patent recites as follows:

28
“wherein said selecting step comprises selecting (3:1) three
consecutive pixels (41,42,43) from a first block and three
consecutive pixels (45,46,47) from a neighboring block, wherein
during said calculating step the first offset is based on the respective
pixel values of the two pixels of each block located closest to the
block boundary, and wherein in case abs[first offset] < first threshold
said modifying step (S4) comprises:

modifying (S40) the respective pixel values of the two pixels (41,45)
located next to the block boundary (20) on the basis of the first
offset, and

modifying (S41) the respective pixel values of the two pixels (42,46)
located one pixel away from the block boundary (20) on the basis of
a second offset and a third offset, respectively, wherein the second
offset is based on the respective pixel values of said three pixels
(41,42,43) of the first block and the first offset, and the third offset is
based on the respective pixel values of the three pixels (45,46,47)
of the neighboring block and the first offset.”

147
Claim 3 of the Patent differs from claim 2 of the Patent in that three
consecutive pixels are selected from a first block and three further
consecutive pixels (45,46,47) are selected from a neighboring block.

148
Cited document D3 discloses, or at least renders obvious, the following
feature of claim 3 of the Patent:

“wherein said selecting step comprises selecting (3:1) three


consecutive pixels (41,42,43) from a first block and three
consecutive pixels (45,46,47) from a neighboring block”

149
This is for instance disclosed in cited document D3. Claim 10 of cited
document D3 recites as follows:

“10. The method according to claim 6, wherein the adjusting step in the
second mode satisfies at least one of the following conditions:

v3'=v3 -d;

v4'=v4 +d;

v2'=v2 -d2;

v5'=v5 +d2;

v1'=v1 -d3 ; and

29
v6'=v6 +d3,

where

d1 =(3(v3 -v4)//8)*δ(|a3,0 |<QP),

d2 =(3(v3 -v4)//16)*δ(|a3,0 |<QP), and

d3 =(3(v3 -v4)//32)*δ(|a3,0 |<QP),

wherein v0 -v7 are initial boundary pixel values, v1 '-v6 ' are
adjusted boundary pixel values, a3,0 is the discontinuous
component of the discrete cosine transform coefficients of the first
pixel and QP is a quantization parameter of a macroblock
containing v4.

150
In summary, claim 10 of cited document D3 clearly teaches three
consecutive pixels v1 -v3 from a first block and three consecutive pixels
v4–v6 from a neighboring block v1 -v3.

151
Cited document D3 discloses, or at least renders obvious, the following
feature of claim 3 of the Patent:

“wherein during said calculating step the first offset is based on


the respective pixel values of the two pixels of each block located
closest to the block boundary”

152
This is for instance disclosed in cited document D3, claim 10 of cited
document D3 defining d1 =(3(v3 -v4)//8)*δ(|a3,0|.

153
Cited document D3 discloses, or at least renders obvious, the following
feature of claim 3 of the Patent:

“and wherein in case abs[first offset] < first threshold said modifying
step (S4)comprises: modifying (S40) the respective pixel values of
the two pixels (41,45) located next to the block boundary (20) on the
basis of the first offset, and

154
The above recited feature of claim 3 of the Patent is for instance disclosed
by claim 10 of cited document D3, i.e., that the pixels v3 and v4 are
modified by the offset d, v3'=v3 -d and v4'=v4 +d.

155
Cited document D3 discloses, or at least renders obvious, the following
feature of claim 3 of the Patent:

30
“modifying (S41) the respective pixel values of the two pixels
(42,46) located one pixel away from the block boundary (20) on the
basis of a second offset and a third offset, respectively”

156
The above recited feature of claim 3 of the Patent is for instance disclosed
by claim 10 of cited document D3.

157
That is, the values of the two pixels located one pixel away from the block
boundary are modified by an amount -d2 or -d3. Claims 10 of cited
document D3 in this regard recites:

v2'=v2 -d2;

v5'=v5 +d2;

v1'=v1 -d3; and

v6'=v6 +d3,

158
Cited document D3 discloses, or at least renders obvious, the following
feature of claim 3 of the Patent:

“wherein the second offset is based on the respective pixel values


of said three pixels (41,42,43) of the first block and the first offset,
and the third offset is based on the respective pixel values of the
three pixels (45,46,47) of the neighboring block and the first offset”

159
The above recited feature of claim 3 of the Patent is for instance disclosed
by claim 10 of cited document D3, which recites as follows for the second
and the third offset:

d2 =(3(v3 -v4)//16)*δ(|a3,0 |<QP), and

d3 =(3(v3 -v4)//32)*δ(|a3,0 |<QP),

160
Therefore, the subject matter of claim 3 is rendered obvious by a
combination of cited documents D3 and D5.

5. Lack of novelty and lack of inventive step of claim 7 with regard to


cited document D3

161 Claim 7 of the Patent is directed to “A deblocking filter unit (60) for
reducing blocking artifacts associated with consecutive pixels of a block
boundary (20) of an image”.

31
162
Basically, the only intrinsic difference between method claim 1 and
apparatus claim 7 is given in feature 7.1, which correspond to a
deblocking filter unit.

163
In other words, the only difference between claims 1 and 7 is given in the
absence of the structural, apparatus-related features in the method claim.

164
Thus, it remains to show that also a deblocking filter unit in terms of a
computer or a computing or data processing unit is disclosed or at least
anticipated by the cited documents.

165 Cited document D3 teaches coding systems as such. In particular, column


1, lines 60 to 62 of cited document D3 teach:

“various methods for reducing a blocking artifact in a coding


system, which individually processes blocks, are presented.”

166
Further, it is respectfully submitted that such systems as claimed by claim
7 of the Patent implementing methods as claimed by claim 1 of the Patent
would also be common general knowledge for the person skilled in the
art.

167
Therefore, claim 7 lacks novelty and inventive step over D3.

6. Lack of novelty and lack of inventive step of claim 22 with regard to


cited document D3

168 Claim 22 of the Patent is directed to “A computer program (74) for


reducing blocking artifacts associated with consecutive pixels of a block
boundary (20) of an image, the computer program (74) comprising code
means which when run on a computer (70) causes the computer to [...]”.

169
The only intrinsic difference between method claim 1 and computer
program claim 22 is given in feature 22.1, which corresponds to a
computer program which when run on a computer causes the computer
to perform the method of claim 1.

170
It is respectfully submitted that such computer programs as claimed by
claim 22 of the Patent implementing methods as claimed by claim 1 of the
Patent would also be common general knowledge for the person skilled
in the art.

171
Therefore, claim 22 lacks inventive step over D3.

32
II. Lack of novelty and inventive step over D4

172
The subject matter of the independent claims of the Patent is disclosed
by or at least rendered to be non-obvious over cited document D4 for the
following reasons.

1. Overview of cited document D4

173
Cited document D45 is directed to “remov[ing] blocking artifacts in low bit-
rate block-based video coding.” See cited document D4, in particular the
Abstract of D4.

174
Accordingly, cited document D4 is in the same technical field or at least in
a neighbouring technical field of the technical field of invention as the
Patent (video coding) and is directed to the same problem to be solved
(reducing blocking artifacts).

175
Cited document D4 is therefore a prior art document in a neighbouring
technical field of the Patent.

176 Cited document D4 states that “[t]raditional block-based video


coders…suffer from annoying blocking artifacts when they are applied in
low bit-rate coding because interblock correlation is lost by block-based
prediction, transformation, and quantization.” See cited document D4,
156.

177 Therefore, cited document D4 proposes a filter that “performs one-


dimensional filtering along the boundaries of an 8x8 block.”

178 Cited document D4 illustrates in Figure 1 (reproduced below) a line of


pixels S1 positioned relative to horizonal and vertical block boundaries.
Cited document D4 notes that the line of pixels S1 “is located across the
block boundary.” See cited document D4, 158.

179
The line of pixels S1 contains two consecutive pixels from a first block,
v3 and v4, and two consecutive pixels from a neighbouring block, v5 and
v6.

180
The first block and the neighbouring block are located on opposite sides
of a block boundary and next to the block boundary. The pixels v3, v4,
v5, and v6 form a line of pixels that is perpendicular to the block
boundary.

5 Two of Kim IEEE’s authors are also co-inventers of Kim. See Section VII.B.1.i,
infra, for a comparison of the disclosures of Kim and Kim IEEE.

33
181
A comparison is given below between the Patent’s FIG. 2a and 2b and
FIG. 1 of cited document D4, each annotated similarly.

182
The depicted comparison shows that v3, v4, v5, and v6 correspond to p1,
p0, q0, and q1, respectively.

The Patent, Fig. 2a

The Patent, Fig. 2b Cited document D4, FIG. 1

183
Cited document D4 explains that “if a four point pixel array S1 is located
across the block boundary, four-point DCT basis vectors of S1 have
symmetric and anti-symmetric properties around the center of four points,
or the block boundary.”

184 See cited document D4, 158. Figure 3 of cited document D4 reproduced
below, illustrates the four DCT basis functions used to calculate
corresponding DCT coefficients for the line of pixels S1. See cited
document D4, 158.

34
Cited document D4, FIG. 3

185
The basis functions are wave functions (wave numbers 0, 1, 2, and 3)
used to calculate corresponding DCT coefficient a0,1, a1,1, a2,1, and a3,1,
where the first subscript denotes the wave number.

186
Cited document D4 notes that the fourth DCT coefficient a3,1 affects block
discontinuity and can be used to reduce blocking artifacts. See cited
document D4, 158; see cited document D4, 156 (“we examine the
existence of the offset in the region rather than the existence of the edge
around the block boundary”); See cited document D4, 157 (“Noticeable
blocking artifacts may result from the concatenation of two flat regions
with a small offset”).

35
187 Based on these observations, cited document D4 proposes a deblocking
scheme with a default mode that is based on a3,1, as illustrated in
Figure 2 below:

Cited document D4, FIG. 2

36
188
The deblocking scheme includes calculating the DCT coefficients a3,k,
(where 0 ≤ k ≤ 2), comparing the absolute value of a3,1 to the quantitative
parameter QP, applying deblocking filtering in a default mode when the
absolute value of a3,1 is less than QP, and applying no deblocking filtering
when the absolute value of a3,1 is not less than QP.

189
The deblocking scheme further discloses applying deblocking filtering in
a smooth region mode, which is used for flat regions of the image. See
cited document D4, 156.

190
In Figure 2 of cited document D4, the DCT coefficients a3,k are calculated
using the following expression:

191
𝑎3,𝑘 = (𝑐1 ∙ 𝑣2𝑘+1 − 𝑐2 ∙ 𝑣2𝑘+2 + 𝑐2 ∙ 𝑣2𝑘+3 − 𝑐1 ∙ 𝑣2𝑘+4 )⁄𝑐3 (10)

192
where 0 ≤ k ≤ 2. For the pixel array S1, k = 1 and equation (10) may be
expressed6 as:

193 𝑎3,1 = (𝑐2 × (𝑣5 − 𝑣4 ) − 𝑐1 × (𝑣6 − 𝑣3 ))⁄𝑐3 (11)

194
As explained above, v3, v4, v5, and v6 denote pixel values for each of the
two consecutive pixels from the first block (v3 and v4) and the two
consecutive pixels of the neighbouring block (v5 and v6) and correspond
to pixel values p1, p0, q0, and q1 of the Patent.

195
In experimental tests using an MPEG-4 video codec, cited document D4
reports that constants c1, c2, and c3 “are approximated to 2, 5, and 8,
respectively, so that the filtering operation may require only integer
multiplication and shift operations.” See cited document D4, 158.
Equation (1) of the Patent is replicated here for comparison:

196
first offset = (9 × (𝑞0 − 𝑝0) − 3 × (𝑞1 − 𝑝1))⁄16 (1)

197
Comparing equation (1) of the Patent to equations (10) and (11) of cited
document D4, equation (1) assigns numerical values 3, 9, and 16 to
constants c1, c2, and c3 and is otherwise identical.

198
Consistent with cited document D4, the Patent assigns integer values to
c1 and c2 (i.e., 3 and 9) and the value assigned to c3 is a multiple of 2 (i.e.,
16) so that only integer multiplication and shift operations are required.
Thus, DCT coefficient a3,1 of cited document D4 corresponds to the
Patent’s first offset.

6Setting k=1 in equation (10) yields:


𝑎3,1 = (𝑐1 ∙ 𝑣3 − 𝑐2 ∙ 𝑣4 + 𝑐2 ∙ 𝑣5 − 𝑐1 ∙ 𝑣6 )⁄𝑐3
Grouping the terms in the numerator by constants c1 and c2 yields equation 11.

37
199 Figure 2 of cited document D4 also shows that a check of |𝒂𝟑,𝟏 | < 𝑸𝑷 is
performed to determine whether to apply the default mode of filtering or
not. See also cited document D4, 158 (“If the magnitude of a3,1 is greater
than a certain value (which is related to the quantization parameter),
however, the filter is not applied to preserve the image details”).

200
This is identical to the Patent’s claimed step of comparing the first offset
to a first threshold.

201 Here, QP corresponds to the claimed first threshold. Cited document D4


notes, in Figure 2, that QP is the quantization parameter of the macroblock
where v5 belongs.

202
In Figure 2, if |𝑎3,1 | < 𝑄𝑃 is false (i.e., “NO” arrow), no filtering is applied
and the pixel values are not modified (i.e., v'=v). Alternatively, if |𝑎3,1 | <
𝑄𝑃 is true (i.e., “YES” arrow), pixel values v4 and v5 are modified/filtered
as follows:

𝑣4′ = 𝑣4 − 𝑑 (12)

𝑣5′ = 𝑣5 + 𝑑 (13)

203
The variable d is a function of a3,1 as shown in Figure 2. In equations (12)
and (13), the value of pixel v4 of the first block is modified by -d (e.g., the
pixel value is decreased if d is positive) and the value of pixel v5 of the
neighbouring block is modified by a corresponding amount with opposite
sign, +d (e.g., the pixel value is increased if d is positive).

204
This reduces the difference between the pixel values on opposite sides of
the block boundary (their modified values are closer together than the
unmodified values) and thus filters the blocking artifact.

205
Cited document D4 further discloses applying filtering in a smooth region
mode. See also cited document D4., Figure 2, 157-58. In that mode,
“strong smoothing is applied inside the block as well as on the block
boundary.” See cited document D4, 156.

206
That is, the values of at least two consecutive pixels on each side of the
block boundary are modified (filtered) in order for the filtering to extend to
the inside of the block. See cited document D4, 156.

207
As in the default mode, in the smooth region mode a check is performed
“[t]o prevent real edges in the filtering region from smoothing.” See cited
document D4, 157.

38
2. Differences between cited document D4 and cited document D3

208
Although two of the authors of cited document D4 are co-inventers of cited
document D3, and some portions of the two references are similar, cited
document D4 is not cumulative of cited document D3.

209
As an initial matter, cited document D4’s deblocking method is different
from cited document D3’s deblocking method. See, e.g., cited document
D4, Fig. 2 (depicting cited document D4 s deblocking method); cf. cited
document D3, Fig. 3 and 4:45-5:50 (describing cited document D3’s
deblocking method, which uses different equations and has a different
structure relative to cited document D4).

210
A particular difference between cited document D4 and cited document
D3 is that, when the absolute value of a3,1 is less than QP, cited document
D4 filters the pixel values v4 and v5 explicitly based on a3,1.

211
That is, a3,1 is used to determine the value of d, which is then used to filter
v4 and v5 in equations (12) and (13). By contrast, cited document D3’s
filtering is not explicitly based on a3,0 when the value of a3,0 is less than
QP. The following excerpt from Fig. 2 of cited document D4 highlights
where a3,1 is used to determine the value of d.

212

213
Cited document D4, Fig. 2 (annotated excerpt).

214
The clipping function CLIP (x,p,q) clips “x” to a value between p and q.
See cited document D4, Figure 2. Analysing these equations, cited
document D4 observes that “[i]f a3,1 increases by 𝜖 we can find that 𝑣4′ =
𝑐
𝑣4 − 𝑘1 ∙ 𝜖 and from 𝑣5′ = 𝑣5 + 𝑘1 ∙ 𝜖,” where 𝑘1 = 2. See cited document
𝑐3
D4, 158. Therefore, in cited document D4, the DCT coefficient a3,1
operates as an estimate (scaled by 𝑘1) of the modification of the pixels v4
and v5. Cited document D3’s equations (8) and (9) are reproduced below
for comparison.

𝑑1 = (3 (𝑣3 − 𝑣4 )⁄8) × (𝛿(|𝑎3,0 | < QP)) (8)

39
𝑑2 = (3 (𝑣3 − 𝑣4 )⁄16) × (𝛿(|𝑎3,0 | < QP)) (9)

215
See cited document D4, 4:60-65. Unlike cited document D4, according to
cited document D3, the expressions for d1 and d2 do not include a3,0 when
the absolute value of a3,0 is less than QP—that is, after the delta functions
in equations (8) and (9) have been evaluated.

3. Lack of novelty and lack of inventive step of claim 1 with regard to


cited document D4

216
Claim 1 of the Patent recites as follows:

217
“A method of reducing blocking artifacts associated with consecutive
pixels of a block boundary of a digital image, the method comprising:”

218 Cited document D4 discloses, or renders obvious, the preamble.


Specifically, cited document D4 discloses “a method to remove blocking
artifacts in low bit-rate block-based video coding.” See cited document
D4, 156.

219
Cited document D4 notes that “a video sequence consists of a set of
image frames”, i.e. a set of digital images. Cited document D4 illustrates
in FIG. 1 (reproduced below) consecutive pixels with pixel values v0
through v9 that may be used for filtering on either a vertical boundary/edge
or a horizontal boundary/edge.

40
Cited document D4, FIG. 1

220
Therefore, cited document D4 discloses, or renders obvious, the
preamble.

“selecting (S1) at least two consecutive pixels (41,42) from a first


block and at least two consecutive pixels (45,46) from a
neighboring block, the blocks being located on opposite sides of a
block boundary (20) the consecutive pixels (41,42,45,46) being
located next to the block boundary, and the pixels forming a line of
pixels (41,42, 45,46) which is perpendicular to the block boundary
(20);”

221
Cited document D4 discloses, or renders obvious, this limitation. Cited
document D4 illustrates in Figure 1 (reproduced below) a line of pixels
S1 positioned relative to a horizontal and vertical block boundaries. Cited
document D4 notes that the line of pixels S1 “is located across the block
boundary.” See cited document D4, 158.

222
The line of pixels S1 contains two consecutive pixels from a first block,
v3 and v4, and two consecutive pixels from a neighbouring block, v5 and

41
v6. The first block and the neighbouring block are located on opposite
sides of a block boundary and next to the block boundary.

223
The pixels v3, v4, v5, and v6 form a line of pixels that is perpendicular to
the block boundary. A comparison below between the Patent’s Figures
2a and 2b on the one hand and cited document D4’s Figure 1 on the other
hand, each of which annotated similarly, shows that v3, v4, v5, and v6
correspond to p1, p0, q0, and q1, respectively.

The Patent, Fig. 2a

The Patent, Fig. 2b Cited document D3, FIG. 1

224
Cited document D4 therefore discloses, or renders obvious, this limitation.

“calculating (S2) a first offset based on the pixel values of the four
pixels (41,42,45,46), constituting two pixel pairs, each pixel pair
comprising corresponding pixels from both blocks, by determining
the relative difference between the pixel values of the first pixel
pair (41,45) and the relative difference between the pixel values of
the second pixel pair (42,46);”

225
Cited document D4 discloses, or renders obvious, this limitation. As
discussed above, cited document D4 discloses that the DCT coefficient
a3,1 affects “block discontinuity.” See cited document D4, 158.

42
226
Analogously, the Patent explains that the “first offset” is related to the
“relative difference” between pixel values on either side of a block
boundary. See claim of the Patent, see the Patent, [0028].

227
Accordingly, both DCT coefficient a3,1 of cited document D4 and the
Patent’s “first offset” describe a “discontinuity” or “difference” in pixel
values across the block boundary.

228
Cited document D4 further discloses calculating a3,1 as part of the
deblocking scheme depicted in FIG. 2 of cited document D4. As
discussed in above, cited document D4 discloses calculating the DCT
coefficients (including a3,1) as follows:
𝑎3,𝑘 = (𝑐1 ∙ 𝑣2𝑘+1 − 𝑐2 ∙ 𝑣2𝑘+2 + 𝑐2 ∙ 𝑣2𝑘+3 − 𝑐1 ∙ 𝑣2𝑘+4 )⁄𝑐3 (10)

where 0 ≤ k ≤ 2. Setting k=1 yields the following intermediate equation:

𝑎3,1 = (𝑐1 ∙ 𝑣3 − 𝑐2 ∙ 𝑣4 + 𝑐2 ∙ 𝑣5 − 𝑐1 ∙ 𝑣6 )⁄𝑐3


229
Grouping the terms in the intermediate equation based on constants c1
and c2 yields the following equation
𝑎3,1 = (𝑐2 × (𝑣5 − 𝑣4 ) − 𝑐1 × (𝑣6 − 𝑣3 ))⁄𝑐3 (11)

230 As explained above, v3, v4, v5, and v6 denote pixel values for each of the
two consecutive pixels from the first block (v3 and v4) and the two
consecutive pixels of the neighbouring block (v5 and v6) and correspond
to pixel values p1, p0, q0, and q1 of the Patent.

231 Although claim 1 does not require a particular equation for calculating the
“first offset,” the Patent’s equation (1) is reproduced below for comparison:

first offset = (9 × (𝑞0 − 𝑝0) − 3 × (𝑞1 − 𝑝1))⁄16 (1)

232
Thus, a3,1 corresponds to the claimed first offset. Moreover, a3,1 is
calculated as a function of pixel values of two consecutive pixels from the
first block and two consecutive pixels of the neighbouring block.

233
Cited document D4 therefore discloses, or renders obvious, the above
recited limitation.

234
Further, cited document D4 also discloses, or at least renders obvious,
the following limitation:

43
“comparing (S3) the first offset to a first threshold value, and”

235
For example, cited document D4 discloses comparing the DCT coefficient
a3,1 (first offset) to a quantization parameter QP (first threshold) by
evaluating the expression |𝑎3,1 | < 𝑄𝑃. See cited document D4, Figure 2;

236
See also cited document D4, 158 (“If the magnitude of a3,1 is greater than
a certain value (which is related to the quantization parameter), however,
the filter is not applied to preserve the image details”).

237
Analogously, the Patent explains that “the first threshold is determined on
the basis of quantization parameter values assigned to the respective
block.” See the Patent, [0100].

238 Thus, quantization parameter QP of cited document D4 corresponds to


the Patent’s first threshold, cited document D2 paragraphs 116-118, and
cited document D4 discloses comparing the DCT coefficient a3,1 (first
offset) to the quantization parameter QP (first threshold).

239 Therefore, cited document D4 discloses, or renders obvious, this


limitation.

“modifying (S4) the respective pixel values of the first pixel pair
(41,45) , by adding said first offset to one of said consecutive pixels
(41) and by subtracting said first offset from the other of said
consecutive pixels (45), in case abs[first offset] < first threshold, or
not modifying any pixel at all, in case abs[first offset] >= first
threshold.”

240
Cited document D4 renders this limitation obvious. For example, as noted
above, cited document D4 discloses, in Figure 2, that, if |𝑎3,1 | < 𝑄𝑃 is true
(“YES” arrow), pixel value v4 from the first block and pixel value v5 from
the neighboring block are modified as follows:

𝑣4′ = 𝑣4 − 𝑑 (12)

𝑣5′ = 𝑣5 + 𝑑 (13)

241
That is, the value of pixel v4 of the first block is modified by an amount -d
(e.g., the pixel value is decreased if d is positive) and the value of pixel v5
of the neighbouring block is modified by a corresponding amount with
opposite sign, +d (e.g., the pixel value is increased if d is positive).

242 This reduces the difference between the pixel values on opposite sides of
the block boundary (i.e., their modified values are closer together than the
unmodified values) and thus filters the blocking artefact.

44
243 Otherwise, if |𝑎3,1 | < 𝑄𝑃 is false (“NO” arrow), the pixel values are not
modified (i.e., v'=v). See cited document D4, Fig. 2, see also cited
document D4, 158 (“If the magnitude of a3,1 is greater than a certain
value (which is related to the quantization parameter), however, the filter
is not applied to preserve the image details”).

244
The pixel modifications of equations (12) and (13) are applied in the
“default mode” of cited document D4. See cited document D4, Figure 2.
Accordingly, cited document D4’s “default mode” modifies the pixel value
of one pixel from the first block (v4) and the pixel value of one pixel from
the neighboring block (v5) when |𝑎3,1 | < 𝑄𝑃 is true. The “default mode”
does not, however, modify the pixel values of two consecutive pixels
from the first block and the pixel values of two consecutive pixels from
the neighboring block, as claimed.

245
However, it would have been obvious to “modify…two consecutive pixels”
from each block in cited document D4’s “default mode,” as claimed.

246 For example, cited document D4 suggests that variations of the “default
mode” of Figure 2 can include applying smoothing to “a few pixels around
the block boundary” (a person skilled in the art would have understood
this to mean more than a pair of pixels directly along the block
boundary). See cited document D4, 157, see cited document D2, ¶83.

247
As further discussed below, a person skilled in the art would have been
motivated to modify two or more consecutive pixels on each side of the
block boundary (i.e., “a few pixels”) in the “default mode” and would have
had a reasonable expectation of success in doing so.

248
Therefore, cited document D4 renders this limitation obvious.

249
Because cited document D4 discloses or renders obvious all the
limitations of claim 1, cited document D4 renders claim 1 obvious.

4. Motivation to “Modify…Two Consecutive Pixels” in the “Default


Mode”

250
At first, as depicted in Figure 2, cited document D4 discloses applying
filtering to a single pixel from each block in the “default mode.”

251
Further, besides the default mode, cited document D4 also additionally
suggests applying filtering to two or more consecutive pixels from each
block, and a person skilled in the art would have recognized the benefits
of doing so.

45
252
For example, when describing the “default mode” cited document D4
states that “smoothing of a few pixels around the block boundary is
enough to achieve a desired deblocking effect” See cited document D4,
157.

253
A person skilled in the art would have understood that, in this context, a
“few” means more than a pair of pixels directly along the block boundary.
See cited document D2, ¶83.

254
Cited document D4 notes that modifying a single pixel along the block
boundary can be sufficient in certain implementations, such as that of
Figure 2 (see e.g., cited document D4, 157), but by no means restricts
filtering to a single pixel on each side of the block boundary.

255 This understanding that cited document D4 suggests modifying more than
one pixel in each block is reinforced in view of related teachings by the
co-authors of cited document D4.

256
Specifically, in cited document D3, when describing a similar deblocking
filtering scheme (notwithstanding the different equations), the cited
document D4 co-authors disclosed modifying two consecutive pixels on
each side of the block boundary.

257
The remaining features of cited document D3’s deblocking method closely
correspond to cited document D4’s “default mode”.

258 Thus, the cited document D4 co-authors were motivated to and were
successful in implementing a filtering method similar to the “default mode”
where two consecutive pixels on each side of the block boundary were
modified. See cited document D2, ¶¶84-85.

259 Even the Patent acknowledged that it was known to modify two
consecutive pixels on each side of the block boundary when applying
filtering in contexts similar to cited document D4 “default mode”.

260 Applying these formulas to two consecutive pixels from each block in
Cited document D4’s “default mode” would have yielded expected results
in accordance with the mathematical expressions provided.

261
Similarly, cited document D4 discloses formulas for modifying at least two
consecutive pixels on each side of the block boundary (see cited
document D4, Figure 2, box labeled “smooth region mode”) and shows
exemplary results of performing filtering according to these formulas (see
cited document D4, Figures 4-6).

46
262
Cited document D4 notes that applying these formulas “provide[s] a
strong smoothing effect inside a block as well as on the block boundaries”.
See cited document D4, 157.

263
Cited document D4 further provides detailed performance evaluations of
the proposed deblocking filter in Tables I and II. See cited document D4,
159, 160.

264
Thus, a person skilled in the art would have understood the expected
impact of filtering two consecutive pixels from each block on both the
subjective appearance of filtered images (i.e., providing a stronger
smoothing effect than filtering only one pixel) and the objective
performance of the filter.

265 A person skilled in the art would further have been motivated to modify
two consecutive pixels on each side of the block boundary in cited
document D4’s “default mode” because it would have amounted to the
simple substitution of one known element (modifying two consecutive
pixels from each block) for another (modifying one pixel from each block)
to obtain predictable results (providing a stronger smoothing effect than
filtering only one pixel).

266
This substitution would have been readily achievable by a person skilled
in the art via simply editing the software of cited document D4’s deblocking
filter, and would have provided the predictable result of increasing the
strength of the deblocking filter.

267
Cited document D4 does not teach away from modifying more than one
pixel on each side of the block boundary in the “default mode.” On the
contrary, as explained previously, cited document D4 suggests that in
various implementations, more than one pixel on each side of the block
boundary can be modified. See cited document D4, 157.

268 At most, cited document D4 cautions that applying too strong of a filter
can introduce “undesired blur”. See cited document D4, 156.

269
However, this undesired blur is mitigated by “sophisticated smoothing
operation” in the “default mode” and not by restricting filtering to only one
pixel on each side of the block boundary.

270
Thus, a person of skill in the art would find it obvious to modify two
consecutive pixels on each side of the block boundary in the “default
mode” when considering cited document D4.

47
271
Accordingly, cited document D4 renders this limitation obvious, and a
person skilled in the art would have been motivated to “modify…two
consecutive pixels” in the “default mode” of cited document D4.

272
Thus, in summary, the subject matter of each of claim 1 of the Patent fails
to meet the requirements of sections 72(1)(c), 14(3) of the 1977 Act.

D. Summary and Conclusion

1 It has been shown hereinabove that each of the independent claims lacks
novelty over numerous prior art documents.

2
Therefore, the Patent is to be revoked in its entirety under section 72(1)(a)
of the 1977 Act.

3
In addition, it has further been shown hereinabove that each of the
independent claims lacks inventive step over numerous prior art
documents or a combination thereof.

4
Therefore, the Patent is to be revoked in its entirety under section 72(1)(c)
of the 1977 Act.

Maiwald Patentanwalts- und Rechtsanwaltsgesellschaft mbH

Dr. Simon Q. Lud Christopher Holtby


Patentanwalt | European Patent Attorney UK Patent Attorney | European
Patent Attorney

Enclosures
Annex M1 (list of references);
Cited documents D1 to D17

48

You might also like