一种方向提升小波变换和网格编码量化的图像编码算法

An Image Coding Algorithm Combining Directional Lifting Wavelet Transform and Trellis Coded Quantization

由于JPEG2000图像压缩标准不能有效表示图像边缘的缺陷,提出了一种结合方向提升小波变换和网格编码量化(TCQ)的图像编码算法.该算法先通过基于率失真优化的四叉树分割算法得到最优变换方向,再沿最优变换方向做方向提升小波变换,然后对得到的小波子带系数进行TCQ量化,最后将量化系数编码输出.实验结果表明,对所给定的测试图像,该算法比JPEG2000峰值信噪比最多提高了1.49 dB,比JPEG2000的结构相似度最多提高了3.98%,解码图像的主观视觉质量更好.

An image coding algorithm combining the directional lifting wavelet transform （DL- WT） and the trellis coded quantization （TCQ） is presented to overcome the drawback of JPEG2000 image compression standard in representing image edges. This algorithm firstly uses a rate-distortion optimization-based quadtree segmentation method to get the optimal transform directions, performs DLWT along the optimal transform directions, then quantizes the wavelet subband coefficients by TCQ, and finally encodes the quantization coefficients outputting the bitstream. Experimental results show that the proposed image coding algorithm outperforms JPEG2000 for the given test images up to 1.49 dB in peak signal-to-noise ratio and 3.98% relatively in multi-scale structural similarity index, and that significant improvement in the subjective visual quality of decoded images is also observed.