You are on page 1of 21

HARDWARE IMPLEMENTATION

OF TRACE TRANSFORM
PRAMOD K
(USN: 2BV09LDE06)

UNDER GUIDENCE OF
Prof. MEENA S M
CONTENTS
• INTRODUCTION
• TRACE TRANSFORM ALGORITHM
• PROPOSED DESIGN
• WORK IN 1ST SEMESTER
• WORK IN 2ND SEMESTER
• WORK IN PROGRESS
• FUTURE WORK.
INTRODUCTION
NEED FOR IDENTIFICATION OF AN IMAGE

• Databases store images. Ex: Web.


• Image transfer through networks.
– Mails.
– Online Medical Support.
• Image encryption and decryption for security purposes.
– Sender.
– Receiver.
• Hence, Identify a given image with the reference image for
modifications.

• HOW TO IDENTIFY???
IMAGE IDENTIFIERS
• Several ways of Image Identification.
– Image similarity.
– Watermarking.
– Image Identifiers.
• Image Identifiers.
– Capture the features of an image.
– Sequence of 0’s & 1’s.
– Properties
• Unique.
• Robust.
• Compact.
• Easy for extraction.
• Fast searching.
IMAGE TRANSFORMATIONS

• Why Transformation Required?


– Spatial domain to alternate domain.
– Easy to extract the information, Identifiers.

• Types of Transformations.
– Radon Transformation.
– Trace Transformation.
TRACE TRANSFORMATION
• 1998, Kadyrov & Petrou.
• Generalization of Radon Transformation.
• Functionals are applied over the defined trace
lines.
• For every θ & every d, the functional is
calculated.
– Output: 1D Binary Identifier.
IMAGE IDENTIFIER EXTRACTION
ALGORITHM

• 2 stages
– Pre-processing stage.
– Processing stage.
• Pre-processing Stage.
– Resize the image to NxN, convert to grey scale.
– Extract the left corner region from the resized image.
– Filter the image with Gaussian kernel.
IMAGE IDENTIFIER EXTRACTION
ALGORITHM
• Processing stage.
– Take Trace Transform of the image
– ∫f(t) dt [integrating over all lines in the image].
– Apply 2 diametrical functions to columns of 2D matrix resulting
from above step.
– ∫|g(t)’| dt [gradient].
– Max(g(t)).
– Take Magnitude of the F T of 2 circus function.
– Binary strings:
• Difference of neighboring coefficients
– C(w)= |F(W)|- |F(w+1)|,
• Applying threshold
– Bw =
ALGORITHM FLOW

IMAGE

TRACE
TRACE TRANSFORM
TRANSFORM

DIAMETRIC DIAMETRIC
TRANSFORMATION TRANSFORMATION

IDENTIFIER IDENTIFIER
EXTRACTION EXTRACTION

IDENTIFIER
CONCATENATION
TRACE TRANSFORM FLOW

Conversion to grey
Input Trace Transform
scale
(1st Functional)

Out
put

Circus Function
1D Function of
ᶿ Apply Functional to
the columns
Output Input
1 0 1 1 0 1 (diametric
functional)

2D Function
CONCATENATION OF IDENTIFIERS
IDENTIFIER FROM 1ST TRANSFORM IDENTIFIER FROM 2nd TRANSFORM

0 1 1 … 0 0 1 1 … 0


0 1 1 1 … 1 1 0
• DETAILED STUDY OF FPGA DESIGN FLOW IN
XILINX ISE 9.2i
• RTL CODING USING VHDL TO DESIGN TRACE
UNIT FOR AN IMAGE.
• SIMULATED THE TRACE FOR IMAGE FOR AN
ANGLE OF 45 AND 135 DEG.
• SYNTHESISED THE DESIGN FOR PROCESSING 4
IMAGES CONCURRENTLY AT A TIME.
RESULTS
• Device utilization summary:
• ---------------------------

• Selected Device : 5vlx30ff324-3

• Slice Logic Utilization:


• Number of Slice Registers: 62 out of 19200 0%
• Number of Slice LUTs: 109 out of 19200 0%
• Number used as Logic: 109 out of 19200 0%

• Slice Logic Distribution:


• Number of Bit Slices used: 111
• Number with an unused Flip Flop 49 out of 111 44%
• Number with an unused LUT: 2 out of 111 1%
• Number of fully used Bit Slices: 60 out of 111 54%

• IO Utilization:
• Number of IOs: 10
• Number of bonded IOBs: 10 out of 220 4%

• Specific Feature Utilization:


• Number of Block RAM/FIFO: 1 out of 32 3%
• Number using Block RAM only: 1
• Number of BUFG/BUFGCTRLs: 2 out of 32 6%
• PRE CALCULATION OF ADDRESSES USING XMIN
AND XMAX FOR ALL IMAGES ROTATED FROM 0
TO 90 DEG. - Completed
• USING C PROGRAM ADDRESSES IN BINARY ARE
GENERATED.- Completed.
• TRACE COLUMNS FOR 4 IMAGES WITH 0, 5, 10
& 15 DEG ROTATION IS COMPUTED PARALLELY.
• SYNTHESIS REPORT….
HDL Synthesis Report
• Macro Statistics
• # RAMs :8
• 794x14-bit dual-port RAM :8
• # Counters : 18
• 14-bit up counter :4
• 32-bit up counter : 14
• # Registers : 13
• 1-bit register :1
• 14-bit register :8
• 2-bit register :4
• # Latches :2
• 1-bit latch :2
• # Comparators : 16
• 14-bit comparator greater :4
• 14-bit comparator lessequal :4
• 14-bit comparator not equal :4
• 32-bit comparator greatequal :4
FUTURE WORK
• COMPUTE THE TRACE FOR OTHER ROTATIONS
i.e. 0 – 180
• OBTAIN THE BIT STRING FOR THE TEST IMAGE
“LENA”.
• COMPARE IT WITH THE BIT STRING OBTAINED
WITH MATLAB.
THANK YOU

You might also like