Abstract:

A blind restoration algorithm based on blur detection and segmentation of the single image is proposed for the spatial changed motion blurred image. Firstly, in order to generate the corresponding fuzzy area map diagram, the fuzzy region of different space of the blurred image is detected. Secondly, the erosion and dilation methods are used to dispose the fuzzy area and the clear area of the map diagram. Thus the corresponding maps obtained. Then the corresponding blurred image is marked and used to split the fuzzy image by the segmentation algorithm. Furthermore, the blurring kernel of the fuzzy region is estimated by two steps. Then the fuzzy region is expanded and the blurred kernel is used to deblurring by deconvolution. Finally, in order to obtain a complete deblurring image, the extension portion is removed and spliced with a clear divisional regional. The experimental results show that the peak signal to noise ratio (PSNR) improves 10 dB and the root mean square error (RMSE) reduces 0.12 dB. For the real fuzzy data set, the results of visual quality improve obviously in the natural scene. Meanwhile, the complexity of the model is reduced with the proposed algorithm. Also, the quality of the deblurring image is improved.