Research On The Algorithm Of Motion-blurred Images

Image restoration is the process of transforming the input degraded images to the latent ones.The main reason that causes the degradation is due to the environmental noise,camera shake and object moving.Image restoration is one of the most important topics in image processing,and it is applied broadly in industry,security surveillance system,military affairs,medical treatment and so on.These factors can be represented by a blur kernel called PSF(point spread function),and this problem can be categorized into two sections.One is the non-blind deconvolution which means PSF is known,such as defocusing blurring and uniform linear motion blurring.Another one is the blind deconvolution and the target is to estimate both latent image and the blur kernel only given by the input motion-blurred image.In this paper,I introduce a fast,effective blind deconvolution algorithm through a single motion blurred image.On the analysis of a large amount of natural scene image gradient distribution,it's heavy-tailed and sparse.The motion blur kernel is also sparse.Based on the two characteristics,I estimate the blur kernel,then restore the blur image.In this paper,I introduce the bi-Lp-norm algorithm to constrain image gradients and motion kernel,which can restore the salient edges and texture as well as suppress ringing.I use the maximum a posterior(MAP)framework and the split-Bregman method to construct and optimize the energy function.I use the image pyramid to get the iteration solution in the scale space.Besides,we introduce a hysteresis thresholding to suppress the noise in the motion blur kernel.The experimental results show that our algorithm can restore high quality images and suppress ringing effect,which can compete with the state-of-the-art methods.