Research On Sparse Representation-Based Medical Image Compression

In the past few years, with the development of medical imaging technology, a variety of medical imaging equipment produced a large number of medical digital images. Medical digital images have the characteristics of rich content and intuitive presentation, and they can effectively assist medical diagnosis. However, because the amount of data is very large, and it will consume much storage space and a lot of bandwidth in the transmission, it is necessary to compress the medical digital images.Most of the existing image compression standards are based on the orthogonal transformation of image, led by JPEG compression standard, which is based on the discrete cosine transform. JPEG algorithm can achieve excellent compression performance for all types of images. However, JPEG algorithm uses fixed dictionary encoding and decoding. When it is used to compress medical digital images, it does not take into account the characteristics of the medical digital images to further improve the compression performance. In addition, the orthogonal transform of an image is not optimal for the image representation, and it could not capture the regularity and contour feature of the image sparsely. Therefore, the further research breakthrough of the image compression lies in the image representation.In recent years, the sparse representation has become a hot topic in the field of image processing. Sparse representation theory suggests that any signal can be decomposed over the over-complete dictionary sparsely, in other words, we can use the linear combination of few atoms to represent the original signal. The sparse decomposition result over the over-complete dictionary for an image is succinct and fit for human visual characteristics intuitively. The outstanding characteristics of the image sparse representation make it a new way to solve the problem of medical digital image compression.In this paper, the main work and contributions are as follows:(1) A compression method of medical image based on fast sparse representation is proposed in this paper. Firstly, this paper describes the traditional image compression method based on sparse representation. Aiming at the redundant computation of residual in the every iteration for images sparse decomposition, this paper uses Batch Orthogonal Matching Pursuit (Batch-OMP) algorithm to decompose images, and it increases the rate of compression. What’s more, this method is used for the compression of medical images. Experiments show that this method can achieve good compression results, and it is superior to the traditional image compression method based on sparse representation and JPEG method in efficiency.(2) Considering the characteristics of medical sequence images, a compression method of medical sequence images based on sparse representation is proposed. Firstly, it removes the similar or identical information in the sequence images by measuring the similarity between images. At the same time, it records the reference indexes of the similar information. Then the reserved image information is decomposed sparsely over the over-complete dictionary, and the sparse coefficient matrix as a secondary dictionary. By the over-complete dictionary, secondary dictionary and the reference indexes, it can reconstruct the original image sequences. Experiments show that the compression result of the proposed method is better than JPEG when the PSNR is higher. Ask for an excellent quality of medical images, which is to ensure that its PSNR is higher, so the proposed method can be well used for the compression of medical sequence images.