Abstract:
we proposed a regularization-based multi-view blind deconvolution method to complete image restoration of blurred images,which are captured by camera array system. This method is conducted with three procedures: firstly, an optical flow algorithm is utilized torealize the registration of adjacent images. Secondly, we use a multichannel blind deconvolution algorithm to estimate the blurs of eachinput blurred images after registration. Finally, according to the estimated blurs, clear images are obtained using non-blind deconvolutionalgorithm. Experiments show that, compared with single image blind deconvolution algorithm, the proposed method is robust to noise.When dealing with the blurred images captured by a camera array system constructed by ourselves, it performs better in terms of NIQE andSSEQ by 4.20% and 55.60% respectively.