An effcient motion deblurring based on FPSF and clustering

A blurry photograph is a type of degradation of image quality. Blur could be modeled with convolutional operation of an image with a blurring kernel, also known as the point spread function (or PSF). Image deconvolution is the process of recovering the unknown image from its blurred version, given a blurring kernel. It is quite time-consuming by using the recursive process to estimate the kernel. The work proposes an approach to the blurred image into a sharp image using the intelligent computing integrating with linear image degradation process, Fourier transforms, and Fourier spectrum. Based on the model built from the information of the blurry image, a linear degradation process, we estimate the kernel power spectrum and compute phase-retrieval applied the intelligent computing for Fourier theorem and Wiener-Khinchin theorem. Then, the optimal blur kernel can be estimated by using kernel clustering and kernel integration under fast point spread function (FPSF). Finally, the sharp image is achieved by using a deconvolution process, inverse Fourier transform. The approach to deblurring is applied the intelligent computing on the estimated image and Peak signal-to-noise ratio (PSNR) to evaluate the performance. By rebuild an improved PSF, the computing strategy leads a kernel estimation of the caught image and reduces the computational time. Experimental results demonstrate that the proposed method with intelligent computing applied can decrease computational time and achieve good visual quality for deblurring images.