This project introduces a new algorithm for the lossless
distributed source coding of
pairs of still images with high intra and inter pixel correlation.
The algorithm is
based on placing the differences between successive pixel values in each row
of one of the images into cosets, with each coset containing sub-cosets of
that are separated by at least a pre-determined distance. The pre-determined
distance corresponds to the maximum allowable gray level value difference
between the same-location pixels of the two images.
carrying the non-redundant information about difference values, are then
being converted into gray code and being bit-plane coded.
The decoding is done using the other image
ready at the decoder, through an anti-modulus taking operation. Even at
the pre-determined distance of 128, which corresponds to the case that the
allowable distance between the same-location pixels of the two images can be
as large as +/- 64, the algorithm achieves a bit rate very close to that of
the joint entropy of the two images. The power of the algorithm comes from
that it takes into account both the intra-pixel (between the pixels of the
same image) and the inter-pixel (between the same-location pixels of the
two images) correlation.
PROBLEM DESCRIPTION and PRIOR WORK