| In recent years, with the popularization of Multi-Media, the development of Digital video and the spreading of Internet pics, digital image and video become more and more common in all aspect of life. For sake of using bandwidth more effectively and saving the storage media, it is necessary to make lossy compression on image and video. Quantization is one of the key step of Lossy Compression, the performance of quantization has great influence on the image compression coding system.Trellis Coded Quantization (TCQ) is based on the ideas of an expanded signal set, set partitioning and state transfer from Trellis Coded Modulation (TCM). TCQ is a quantization method with both excellent performance and moderate complexity. It has been shown to be an efficient method for memory-less source and Gauss-Markov source.Now, there has been a amount of activity in designing TCQ image coder. Different with the present methods, we want to design a modified TCQ which uses variable trellis against different signal sources instead of a stable one. The modified TCQ uses the information of the data that has been processed to optimize trellis structure to decrease MSE and improve the performance.Based on this idea, we proposed a modified TCQ named Statistics of Residual Based Trellis modification method. It follows the structure of MS-TCQ, adaptively modifies the trellis during the quantization. By using this method to quantize different source, and analyze the performance under several parameters, we get a conclusion that the proposed method works better for both memory-less and memory source.Combine the proposed method with JPEG2000, and set the parameters according to JPEG2000 Standard, we get an algorithm called Adaptive Trellis Coded Quantization (ADTCQ). We use 2 different methods to do JPEG2000 compression with a series of testing images. Experiment results show that by using proposed ADTCQ in JPEG2000 image compression has better PSNR performance than using TCQ, and also keeps Visual features. |
PDF Full Text Request