Effect of coefficient coding on JPEG baseline image compression

Summary form only given. This work uses the example quantization table described in the JPEG drafts as the basis for comparing the amount of compression attributable to each of several distinct lossless compression techniques. It takes the JPEG algorithm apart into its constituent pieces. The result shows that most of the JPEG compression performance is achieved from coefficient quantization that replaces coefficient values comprising many bits with values of few bits. Additional compression is achieved by a combination of runlength coding, predictive coding, and either Huffman or arithmetic coding.>