Algorithm Degraded Image Of Variational Bayesian High Resolution

The image is a very important visual source for machine vision system. However, in the image acquisition and imaging system, there usually exists various degradations including motion warping, system blurring and noise effect. These degradations will not only cause the visual quality decline, but also affect the usage of the valid image information. For instance, in the case of performing segmentation of degradation image, the accuracy of segmentation has a serious decline, and thus affects the target detection, recognition and understanding. Therefore preprocessing is an important scientific problem in image processing and computer vision.This article mainly research layered Bayesian super-resolution pretreatment of degradation image about the relevant algorithm. Using hierarchical Bayesian theory, we design and implemented a variational Bayesian super-resolution method based on horizontal and vertical gradient L1sparsity prior. Through the modeling of image degradation, image prior and hyper-parameters prior, we establish a multi-images variational super-resolution model, and we then implemented two kinds of image super-resolution corresponding blind and non-blind mode. Furthermore, using the YUV color space, the algorithms are extended to deal with the color images. Experimental results show that our method can obtain a good estimation for the registration parameters among multiple images and achieve a good super-resolved image.