Researches On Recursive Block Prediction For Image Compression Algorithm Based On Texture Direction

As images become the most intuitive and most important carrier of information interaction,it is important to study better image compression algorithms.In recent years,many scholars have made very good results on the research of block-based intra-frame prediction algorithms,and also pointed out a new direction for the study of still image compression.However,there are still some shortcomings in the intra-frame prediction technology,such as complex selection of prediction mode,ignoring the potentially correlation information in other neighbor pixels of the block and so on.The excellent prediction model is very important to improve the prediction accuracy.At the same time,the accurate detection of texture direction is helpful to select the prediction mode and improve the prediction accuracy.In this paper,we focus on the algorithm of image compression based on prediction and transformation,we focus on the research of recursive block prediction image compression algorithm based on the texture direction.Through the study of texture direction detection method and recursive prediction model,we propose three kinds of recursive predictive image compression algorithms which combine recursive prediction approach with the detection of texture direction after researching on the method of detecting texture direction and the model of recursive prediction.The main contents of this paper include the following aspects:(1)Aiming at the problems that reference pixels of the prediction model are large and the selection of prediction mode is complex,an improved algorithm of recursive block prediction image compression based on optimal direction is proposed.The algorithm previously detects the texture direction of the 4×4 block through directional filter banks of the frequency domain or the statistics of gradient direction histogram,and then determines the prediction mode according to the texture direction.After that,it uses the current block pixels to establish the recursive prediction model of the determined prediction mode to complete the recursive prediction for the current block,until the completion of compression coding for the entire image.The experimental results show that compared with the intra-frame prediction algorithm without considering the rate distortion optimization,the algorithm takes the potential correlation between the neighbor pixels into account,and improves the prediction accuracy by using only a few adjacent pixels while fast deciding prediction mode.At the same time,the accuracy of the texture direction detection approach based on directional filter banks of the frequency domain is higher than that based on the statistics of gradient direction histogram.(2)To utilize the similarity of image texture direction,this paper proposes an improved recursive block prediction image compression algorithm based on Quad-tree Merging which further optimize the new algorithm of recursive block prediction image compression based on optimal direction.The experimental results show that this method establishes a recursive prediction model and estimates the model coefficients for the local merged blocks which possessing same texture direction,thus reduces the auxiliary information of the prediction while almost not affecting prediction performance.(3)In order to make futher use of the similarity of image texture direction,a recursive block prediction image compression algorithm based on estimating coefficients using adjacent block is proposed.The algorithm statistically analyzes the texture relationship between the current block and the adjacent block,and the adjacent block with the highest similarity of texture direction is selected to estimate the prediction coefficients which are used for recursive prediction.The experimental results show that while the auxiliary information is less,the proposed algorithm can improve the coding performance for images with high localized texture similarity.