A Unified LZ and Hybrid Coding for Compound Image Partial-Lossless Compression

This paper proposes a Unified LZ and Hybrid Coding (ULHC) method for compound image and video compression. The method is macroblock-based. Each macroblock is first coded simultaneously by both H.264 and gzip, then rate-distortion optimization is used to select either H.264 coding or gzip coding as the final coding of the macroblock. To improve the compression performance of ULHC, gzip uses the most recent 32k bytes of reconstructed macroblock data as the dictionary. Moreover, deblocking is disabled if at least one macroblock in an image is coded by gzip. Experimental results show that for images composed of natural video picture, text, and graphics, the proposed method provides up to 18 dB higher PSNR than H.264, while for text and graphics mixed images, the rate-distortion performance improvement is more significant. Another important feature is that ULHC can achieve partial-lossless and partial-near-lossless coding with a very high compression ratio, typically over 18.