Study On Image Deblurring System

In the process of image acquisition,the relative motion of the photosensitive element and the measured object,the unreasonable exposure,the camera defocus will cause the degradation of the image,and the degradation due to the relative motion is called motion blur.Optical Stabilizer and mechanical gyroscope can effectively reduce the degradation degree,but because of its high cost and high energy consumption,it can not be applied to the actual monitoring environment.With the leap forward in hardware performance,edge computing has gradually become the development trend in the future,and the front-end image deblurring system also has a certain degree of feasibility.Image deblurring is to adopt certain mathematics means,construct a mathematical model according to the qualitative change principle and statistical features of the image,and recover the target image from the qualitative image by solving the model,so as to achieve the purpose of image stabilization.However,in actual situations,an accurate blurring kernel is usually not available.This becomes a blind deblurring process.Its unknown quantity is larger than the known quantity,and there is no unique solution.It is a serious ill-posed problem and it is extremely difficult to solve.Therefore,research and hardware implementation of image deblurring algorithm have great significance and application scenarios.This paper mainly focuses on the algorithm and implementation of single image deblurring.The main work includes the following aspects:(1)In this paper,the frequency domain characteristics and refinement methods of uniform motion blur are analyzed,and a high-performance point spread function measurement method is proposed.The motion blurring direction measurement is based on the second Fourier,combined with the RANSAC sampling method,effectively improving the accuracy and speed of the blurring angle measurement.(2)This paper analyzes the sparse characteristics of Extreme channels in natural images and the limitations of Extreme channels.A Extreme channel priori based on image preprocessing is proposed,which increases the priori sparsity and improves the robustness of the algorithm.In order to reduce the time-consuming of the algorithm,this paper converts the expression of the blurring kernel into the FFT solution,avoiding large-scale matrix multiplication.Considering the pixel saturation problem of forest fire,special treatment is performed on the saturated region during image reconstruction,which effectively reduces the ringing effect.(3)The algorithm implementation uses the Intel SoC system solution.With DE1-SOC as the development platform,the SoC system is built in Qsys.The hardware system includes LCD display subsystem,memory drive,FFT subsystem module and so on.The software system uses Linux OS as operating environment and Qt as software framework.Combined with OpenCV image processing library,a GUI software for image acquisition,display and image restoration is developed.In this paper,the effectiveness of the proposed method is proved by MATLAB simulation experiments.Compared with other similar algorithms,there is a certain improvement in the removal of ringing effect and the accuracy of blurring kernel estimation.Finally,the hardware system was built and tested in the actual project.The results show that the image deblurring system proposed in this paper has great application value.