Image data compression by DCT with adaptive run-length coding
Shin-ich Arazaki, M. Saigusa, Sakura Hashiguchi, Makoto Ohki, Masaru Uchiyama, F. Itoh · IEEE Transactions on Consumer Electronics · 1991
Efficient adaptive discrete cosine transform (DCT) encoding for data compression is described. This algorithm is simple and is suitable for digital still cameras. At a bit rate of 1.2 b/pixel, a signal-to-noise (S/N) ratio of about 40 dB was achieved when applied to the Standard Image Database (SIDBA) 'girl Y' images. The data for the DCT coefficients are compressed by adaptive run-length coding, in which either the run-length coding or the non-run-length coding was selected based on the number of zero data in each of the subdivided pictures. By adopting nonlinear quantizers, the Huffman code table was successfully compressed with very little degradation in the images.>