Optimal Huffman coding of DCT blocks

It is a well-observed characteristic that, when a discrete cosine transform block is traversed in the zigzag order, ac coefficients generally decrease in size and the runs of zero coefficients increase in length. This paper presents a minor modification to the Huffman coding of the JPEG baseline compression algorithm to exploit this characteristic. During the run-length coding, instead of pairing a nonzero ac coefficient with the run-length of the preceding zero coefficients, our encoder pairs it with the run-length of subsequent zeros. This small change makes it possible for our codec to code a pair using a separate Huffman code table optimized for the position of the nonzero coefficient denoted by the pair. These position-dependent code tables can be encoded efficiently without incurring a sizable overhead. Experimental results show that our encoder produces a further reduction in the ac coefficient Huffman code size by about 10%-15%.