Research On Image Quality Improvement Technology Based On The Blind Restoration

High quality images have become more important in the field of national defense security,deep space exploration,biological medicine,safe city.However,due to the limitations of the production process and cost,the improvement of the hardware can not meet the high requirements of the image definition.Therefore,in the condition that the imaging system equipment does not change,the image restoration technologies to improve image quality have been paid more and more attention because of their adventages of low cost,high practicability and wide application.Because of the complexity and the uncontrollable nature of the shooting environment,such as relative motion between the imaging device and the shooting scene,the recorded images by cameras may be out of line with the actual scene,be blurred or have low resolution,which limits the further applications in image recognition and image understanding.Therefore,this paper discusses the problem of image degradation caused by motion blur,involving the global motion blur and local motion blur respectively,by analyzing the basic model of image degradation and establishing the basic idea of image motion restoration,the corresponding algorithms have been proposed to improve the image quality and restore the clear image.The main work of the paper is as follows.(1)In order to solve the global motion blur problem caused by camera shake,some improvements are implemented to the existing edge prediction method,which combines with the rolling guidance filter and new gradient selection mechanism,so the estimation of point spread function is close to the ground truth.In this procedure,the improved non-blind restoration method is used to final recover work,and the processing model of saturated pixels is introduced,so that the ringing effect caused by the high lights region is reduced greatly in the restoration result,and a higher quality recovery result is obtained.(2)This paper presents a region segmentation approach to deal with the problem of local blur in images,which uses the image singular values to detect and identify the blurred region.Then spectral matting method is adopted for image segmentation,by extracting the blur foreground to be processed and composing the image after the restoration of the blurred area,the clear image could be achieved finally.(3)In order to judge the blurry category of the input image,an adaptive discriminant method is designed in this paper.The method uses the area ratio of the blur area to the whole image as the basis of discrimination and establishes the reasonable threshold to automatically determine the blurry type of the input image,and then selects the appropriate restoration method,the validity of the method is verified by the actual test.(4)Comprehensive evaluation with different indexes are carried out to evaluating the global motion blur and local motion blur restoration algorithm proposed in this paper.By comparing with other studies,the results show that the restoration results obtained by the method perform better in both visual effect and objective evaluation indexes,which could accurately estimate the causes of image quality degradation and reduce the ringing effects,demonstrating the its utility and practicability of this method.