The Research On Image Compression Coding Algorithm Based On EBCOT

As a medium of information transmission, the image has always been an indispensable part in the field of communication. Because of the high spatial redundancy, the uncompressed images occupy more storage space. However, in order to improve the image transmission speed and reduce the storage space, as for image itself, image compression becomes only a solution about reduction in the amount of image data. So the image compression algorithm has become a hot research field to enhance image reconstruction performance in image compression. Because of the multi-resolution scalability and good robustness, embedded image compression encoding is one of the direction of the mainstream form of image compression, and EBCOT algorithm is a typical representative of embedded image compression encoding. This thesis improve EBCOT based image compression algorithm from the following two aspects.The image after the wavelet transform will produce the coefficients of different subbands, and the purpose of the quantization algorithm is to ensure image quality as far as possible while reducing the amount of data. According to the characteristics of the local image information, the low-frequency subbands maximum mapping quantization algorithm is proposed. According to the characteristics of the wavelet high frequency coefficients, this algorithm predicts the low frequency subband coefficients, and the low frequency coefficient of the predicted value is divided into the edge area, while the low frequency coefficient of the predicted value is divided into the smooth area. The low frequency coefficients of different regions are quantized by different quantization steps. High frequency subband coefficients are still quantized by the original dead zone quantization method. This algorithm can improve the encoding speed. While preserving the details of the image, the proposed algorithm can improve the quality of the reconstructed image. And the proposed algorithm can enhance about 0.15 more dBs than deadzone quantization.In the rate preallocation stage, the original allocation scheme is based on the sum of the absolute values of the wavelet coefficients and the corresponding subbands. But the idea does not take into account the importance of information content, so the subband rate pre allocation algorithm based on directional gradient weighting at low bitrate is proposed based on this problem. In the improved algorithm, the original image is processed by using the components of the edge detection operator to the original image downsampled after a certain number of times, then the resulting image is processed by absolute value operator and mean operator, so the average value turns out. According to the average value corresponding to each subband, the change tendency of the high frequency subband coefficient is predicted. Weighted processing is based on the original bit rate allocation scheme in accordance with the predicted information. The algorithm can allocate the target bit rates more reasonably. The experimental results show that the proposed algorithm can better predict the wavelet coefficients change trend, and achieve good effect of image reconstruction. At low bit rates(no more than 1bpp), the proposed algorithm in the optimal solution condition can increase by 1 more dBs and 0.6 more dBs than PCRD algorithm and visual optimization weighted algorithm, respectively.