Research On Ulstrosonic Images Compression Algorithm Based On Hypercomplex

Our nation is a vast country where the deployment of the hospital is scattered, while medical level also have significant and regional differences, especially in rural and remote areas, so it is necessary to develop the remote medical service and technology. However, the bandwidth of communication network cannot be capable of increasing unrestrainedly. This needs to compress the remote transmission data so as to improve the efficiency and stability of transmission. It is safe and accurate for the color doppler ultrasonic diagnosis in medical diagnosis such that it is necessary to research on the ultrasonic images compression.The characteristic of medical image compression is that the diagnosis area is nearly loss compressive while the background area is lossy compressive. The hypercomplex algebra is convenient to represent the 3 channels of color images that makes the coefficients of decomposition are sparse and it is helpful to extract significant information combined with transform methods. Empirical mode decomposition (EMD) as a novel data decomposition method can retain textures in the image in the condition of high compression rate. So the goal of the thesis is to research on the effect of doing loss compression to the diagnosis area based on hypercomplex transform and lossy compression to the background based on EMD.Firstly, the paper summarizes hypercomplex computation with the development of the image processing based on it, and the existing compression method systematically. Secondly, the paper realizes the hypercomplex Singular Value Decomposition (SVD) combined with the SVD computation method in matrix analysis, and put it into practice of ultrasonic color images compression. Next the paper simulates the Discrete Cosine Transform compression process in the JPEG standard, and analyses the disadvantages of DCT compression, meanwhile, verify the feasibility of ultrasonic color images compression using hypercomplex DCT. Then, the paper realizes two basic wavelet compression methods and design simple quantization standard to the magnitude-phase representation of hypercomplex wavelet combined with compression indices analysis. Last, the paper introduces EMD compression mode using non-uniform variable samplings considering the local extrema distribution based on the existing coding method of EMD and has made some innovation.