Linear Motion Blurred Image Realization Of Real-time Recovery

This paper presents a general speed to solve the linear displacement without rotation of blurred images of real-time implementation of recovery scheme.The main content of the study is:Firstly, introduced a motion-blurred image restoration purpose and significance of the research, as well as motion blur and image restoration algorithm for image parameters of the situation at home and abroad.Secondly,basic theories are introduced on motion blur image.such as the character of image noise, the basic nature of the image Fourier transform, image degradation model is derived,and factors of image quality evaluation. To enhance the computing speed of FFT(Fast Fourier Transform Algorithm), Used to improve the real discrete Fourier transform (RDFT) algorithm, which is implemented in 1-Dimensiona Wiener filter to restore images degraded by motion or vibration.Then,a modified edge-parameter analysis method of the blur identification(Referred to as:MEPBI) based on Pulse Coupled Neural Networks (PCNN) model is proposed for accurately identify its blurred image the displacement parameter,which is suitable for the identification of the linear motion blur, Combined with the direction of motion blurred images, get a real-time motion blur point spread function(PSF) to restore image.Lastly,in ensuring the quality of image restoration under the premise, Propose a kind of CPU and GPU in coordination with the work pattern, Uses for to realize the movement blurred image real-time restoration. Namely utilize MEPBI algorithm to obtain PSF of the blurred image in CPU, Subsequently in GPU using the improvement RDFT algorithm as 1-Dimensiona Wiener filter engine, Carries on the restoration to the blurred image, The entire plan can complete under the real-time condition.