Rate-distortion Optimization for Image Compression Using Warped Discrete Cosine Transform

Recently, warped discrete cosine transform(WDCT) has been proposed [1, 2], which is the cascade of allpass filters and DCT. The role of the allpass filters is to warp the input frequency contents to be suitable for compression by DCT. The compression algorithm based on the WDCT outperforms the JPEG [5] that is based on the conventional DCT, especially for the images with high frequency contents. In this paper, we improve the performance of the compression algorithm based on the WDCT. The warping parameter of the allpass filter is optimized based on the rate-distortion sense, whereas the previous algorithm considers distortion only. The proposed optimization algorithm controls the rate and distortion by manipulating input signals before DCT, whereas the conventional optimization method [3, 4] controls the R-D by manipulating data after DCT, i.e., the quantization set. The simulation results show that the proposed algorithm shows better performance than the conventional optimized JPEG at high bit rates. Moreover, the proposed algorithm requires lower computational complexity than the quantizer selection method, because only a single parameter is controlled.