The Study On HD PHOTO Feedback Codeing And Image Post-Processing Technology

HD Photo is a new still-image compression algorithm and file format, developed and patented by Microsoft, and now is under consideration in JPEG committee as the third generation international image compression standard. HD Photo can realize both lossless and lossy compression in the same system, it merits in terms of compression ratio and coding quality with occupying much less memory space. HD Photo supports coding of the region of interest, thumbnail extraction. HD Photo is better than JPEG in compression quality while its compression efficiency is the same as JPEG. Based on the HD Photo compression standard, this paper is composed of the following content:First of all, this paper does some further research on PCT and POT transform included in HD Photo, makes implementation of the two points rotation, one-dimensional four points rotation, two-dimensional rotation and one-dimensional four points overlap operation based on lifting, and derivates the formula particularly. Finally, the effect figures are given.Secondly, based on HD Photo compression standard, a method on feedback coding based on Multi-Scale Structural Similarity criterion is proposed. The algorithm calculates the error of reconstruct image and the original image, then uses it to adjust the quant scale used in coding, realizing the adaptive quantization. The experiment results show, compared to the HD Photo's compression algorithm, the proposed scheme improves subjective visual quality of the coded image.At last, HD Photo compression standard has no image post-processing technology; the decoded image will generate block effect in high compression ratio. In order to solve the problem, two algorithms are proposed to eliminate image block effect. One method is based on 2D dual-tree complex wavelet and hard threshold, another method is based on block matching and 3D filtering. The experiment results show that both of the two algorithms can eliminate block effect while maintaining some image details, and can improve the subjective visual quality of the image.