Mahmoud R. El-Sakka, Khaled Hassanein, and Mohamed S. Kamel, "Statistically Optimized JPEG for Image Classes: A Case Study on Financial Document Image Compression", IEEE Canadian Conference on Electrical and Computer Engineering, CCECE'1998, Vol. 1, pp. 173-176, May 1998, Waterloo, Ontario, Canada.
Abstract
This paper describes a novel method of optimizing the JPEG Quantization Table (Q-Table) for classes of images based on the statistical nature of their DCT coefficients. This scheme is used to come up with an optimized JPEG Q-Table for the class of cheque images. The superiority of the optimized Q-Table is then established at high compression rates from a visual image quality point of view, as well as from a cheque amount recognition accuracy point of view.