Research On Millimeter-wave Image Restoration Algorithm And Its Implementation By DSP

Millimeter wave imaging system provides special advantages over the optical imaging, Infra-Red imaging and microwave imaging systems, it is a hot research in the field of millimeter wave in the 90's. However, the images obtained from millimeter wave imaging system exist degradation, which influence the images'quality. This paper starts with the degradation cause of millimeter-wave image, according to the characteristics of millimeter-wave image about low spatial resolution and stacking noise, we study in depth on a variety of image restoration algorithm and its implementation techniques, propose a kind of image denoising algorithm based on wavelet annlysis and super-resolution image reconstruction algorithm based on POCS techniques, which can effectively improve the quality of millimeter-wave images. At the same time, taking advantage of high processing speed of the high-speed digital signal processor (DSP), we study the implementation technology of image restoration algorithm based on DSP development system.With the improvement of wavelet theory, wavelet annlysis has been applied to image denoising successfully, the denoising algorithm based on wavelet annlysis can denoise effectively and keep the edge information simultaneously. Based on the profound analysis on wavelet image denoising, this paper proposes an image denoising method combining median filter, thresholding and wavelet transform, it also discusses the principles and several key issues of the method in detail, and finally implements the algorithm on the hardware platform and provides the experimental results.Image super-resolution restoration is a method that recovers a higher resolution image from the degraded low-resolution images, it can be divided into frequency domain methods and spatial domain methods. Frequency domain methods can only deal with the degradation that only contain the global translational motion, more importantly, they are difficult to include prior knowledge. Spatial methods use general observation models, can utilize prior restraint knowledge effectively, so they have better adaptability and performance. In spatial domain methods, POCS method use the principles of projection and convex sets to realize super-resolution restoration, the algorithm is relatively straightforward, the use of prior knowledge is more simple and flexible, it can maintain the edges and details of images, POCS is one of the most promising super-resolution restoration methods. Based on the profound analysis and comparison on the image super-resolution reconstruction algorithms, this paper focuses on the detailed study of POCS algorithm, analyzing the factors which influence the results of super resolution, such as matching algorithm, search algorithm and the times of iteration, it uses the block matching algorithm based on diamond search algorithm and artificial method of determining the times of iteration, effectively improves the accuracy of motion estimation and reduces the amount of computation. The experimental tests show that this algorithm can get better visual effects, significantly improve the quality of the image and preserve the detail and edge features of the image.