An Enhanced Wavelet Image Codec: SLCCA PLUS
Xiaobo Jiang, Bo Song, Xinhua Zhuang · 2018
High-resolution image compression remains a hot research topic today due to the widespread use of high-definition displays in computers and mobile devices. Conventionally, there are two distinct categories in image compression: 1) Subband coding based on multi-level discrete wavelet transform (DWT); 2) Hybrid coding based on block prediction and discrete cosine transform (DCT). Wavelet image compression has delivered a much higher coding gain over the classical DCT-based JPEG. More recently, through using a delicate yet innovative block prediction scheme before DCT, HEVC Intra coding has delivered an impressive coding gain over JPEG 2000, where the coding gain is measured by the peak signal-to-noise ratio (PSNR) gain at the same bit rate. However, it is hard to integrate a block prediction scheme with wavelet transform. A block prediction error image could be statistically quite different from an ordinary image. In the paper, we focus on developing a wavelet image coder SLCCA Plus, which is a much-enhanced version of SLCCA. The enhancement involves quantization, cluster filtering, sharedboundary-zero handling and context model design for adaptive arithmetic coding. The experiments on standard images of 512×512 are conducted to compare the proposed SLCCA Plus to JPEG 2000 and SLCCA. On average, SLCCA Plus achieves 7% bit saving over JPEG 2000 and 4% over SLCCA.