Uniform Blind Deblurring Based On Edge Information And Image Depth

Image deblur is an important branch of image restoration. It has been widely used in aerospace, medical instrument, computer vision and image processing. The two main directions of image debluring development lie in improving the quality of deblurred image and improving the operation speed. For the former, this paper put forward two methods based on edge information to improve the quality of the resultant image; for the latter, this paper propose a method based on the depth field of the image for image processing patch, thereby accelerating the deblurring process speed. Three methods are briefly described as follows:1. Based on the existing work, a new method based on the external image patch and the bilateral filter is proposed. Differing from the classical one, the proposed method applys a bilateral filter to the last layer of the Pyramid model, which is the same as the input image. The bilateral filter denoises the image, so that the main edge information of the image is preserved to obtain more clear sharp.2. Based on the summary of the previous work, a new method based on the prior external image block and a sparse regularization is explored. In the first method, specific parameters of the bilateral is required by different images. In this new adaptive method, a new algorithm based on the best of the present effect is used in the last layer of the Pyramid model, which makes the iterative process to estimate the correct phase point and finally obtains the fuzzy kernel of the real scene.3. In order to improve the computational efficiency of image deblurring algorithm, this paper tries a partition strategy based on depth image, which is inspired by non-uniform deblurring. The method divides the input image into blocks according to the calculated depth map and deblur these blocks seperatedly. After the image blocking, the matrix dimension of image convolution is reduced, so the computation efficiency is greatly improved.