You are on page 1of 6

Image Compression

Lecture 33
Lossless Predictive Coding
• Eliminates inter-pixel redundancies by extracting and coding only
new information in each pixel.
• New information =difference between actual and predicted value of
the next pixel.
en  f n  pn

f n  en  pn

m 
pn  round    i f n 1 
 i 1 
m 
pn ( x, y )  round    i f ( x, y  1) 
 i 1 
Example
Sequence: 100,102,120,120,120,118,116

Predictor: 𝑝 𝑛 ( 𝑥 )= 𝑓 ( 𝑥 − 1 ) 𝑎𝑛𝑑 𝑝𝑛 ( 0 )= 𝑓 (0) f(n) p(n) e(n) e(n) f(n)


-----------------------------------------------------

100

102

120

120

120

118

116
Delta Modulation
n s(n) s^(n) e(n) e`(n) s`(n)

1 33 33

2 35 33 2 4.5 37.5

3 34 37.5 -3.5 -4.5 33

4 36 33 3 4.5 37.5

5 35 37.5 -2.5 -4.5 33

6 34 33 1 4.5 37.5

7 35 37.5 -2.5 -4.5 33

8 35 33 2 4.5 37.5

9 38 37.5 0.5 4.5 42

10 44 42 2 4.5 46.5

11 50 46.5 3.5 4.5 51

12 59 51 8 4.5 55.5

13 73 55.5 17.5 4.5 60

14 81 60 21 4.5 64.5

You might also like