Blocked Image Compression And Reconstruction Algorithm

Compressed Sensing (CS) theory has broken through the limit of traditional Nyquist sampling theorem and can complete the signal compression while sampling as well as reconstruct the original signal accurately with only a small amount of sampling points. In recent years, CS has become a research hotspot because of its great advantages in data acquisition, data storage and transmission, data analysis and processing, etc. When coming to image signals, Lu Gan et al proposed the Block Compressed Sensing (BCS) theory on the basis of CS. In their theory, the image is divided into individual blocks and be handled respectively so that data storage space is reduced and the time of data sampling and reconstruction is shortened as well. Based on the advantages of BCS, this paper concentrates on how to improve the accuracy of image reconstruction and designs an image encryption system on the basis of chaotic thoughts from the aspect of information security. The research mainly includes:1) The classification of image blocks. Considering that the information content of each image block varies from each other, less samplings are needed as to the blocks with less information which can be reconstructed accurately while more samplings are required as to the blocks with more information. Thus using the same operators in sampling is bound to fuzziness in some details of the image when it is reconstructed. To solve this problem, a new method is put forward in this paper to calculate the information content of each image block at the encoding side. The method integrates entropy estimation and edge detection of each image block and the image blocks are divided into three classes to be sampled adaptively according to the information amount.2) The allocation Strategy of Sampling rates. Because the information distribution is not the same in different images, some distribute evenly which means the information contents in each block are not of great difference, on the contrary, in others the information contents in each block differ from each other to a large extent. For the two types of images above, using the same strategy to allocate the sampling rate can not reflect the image characteristics effectively. Thus these two types of images are defined in this paper by taking the distribution information of each image into account, and then different strategies to allocate the sampling rates are also proposed.3) The improved algorithm in image reconstruction. At the decoding side, the linear reconstruction algorithm has the advantage of fast speed and low algorithm complexity, so that in view of the above-mentioned three kinds of image blocks, three different linear operators are constructed to reproduce each image block. Then an improved iterative thresholding algorithm which is based on different sparse representation is proposed to remove the block effect and the noise when filtering.4) The design of image encryption system. Starting from the angle of image information security, the chaotic encryption thought is used to encrypt the sampling data which is quite different from the whole image pixels, thus a new image encryption system is designed consequently by the integration of CS and the chaos theory.A series of experiments have been performed and the results show that the strategy of allocating sampling rates would contribute to improve the accuracy of image reconstruction; the image reconstruction algorithm would reduce the time of image reconstruction and at the same time improve the exactitude of the reconstructed images; at last, the designed image encryption system on the basis of both the chaos encryption and the CS theories is proved to reduce the encrypted data while realizing double encrypting which results in high image information security.