HW-2-Arithmetic Coding.2020sp
Arithmetic Coding of the Residual Image with quantization. In HW-1, we use the following filter to predict the
pixel values,
f1=
1/3 1/3 0
1/3 0 0
0 0 0
And verified that the predicted image residual indeed follows a GD like distribution, as shown in the Fig 1.
) and c) below
Figure 1.
a) for a predicted e
or image, as shown in Fig.1d, with quantizer q=4, find out what are the unique symbols in
the residual image after quantization, what is the distribution of these symbols ?
Show your matlab code for intra prediction and quantization, as well as residual symbol distribution estimation
Hint:
-1-
) Use SFU Arithmetic coding implementation to generate bit stream, and count the bits-per-pixel, and
compare that with the residual image entropy H1 and original image pixel entropy H0. You may also use the
Python package (torchac), or matlab implementation (arithenco).
Show your code and results:
-2-
-3-
[40pts] 2. For the bi-level image given in Fig. 2, design a line by line context model, i.e, the C(0) and C(1) fo
each line, and use the previous line distribution as context to code the cu
ent line.
Figure 2. Binary Image
-4-