Blind Text Image Deblurring Algorithm Based on Multi-Scale Fusion and Sparse Priors

The goal of blind text image deblurring is to obtain a clean from the given blurry without knowing blur kernel. Sparsity-based methods have been shown their effectiveness in various models. However, kernel estimation based on sparse priors lack consideration for brightness information about kernel, which will affect restoration effect Besides, previous seldom apply both spatial domain and transform information. We propose novel model multi-scale fusion priors. Besides gradient prior latent image, we add high-frequency wavelet coefficients better constrain solution space good images. semi-quadratic splitting method used alternately optimize image. Meanwhile, consider influence feature restored By technique basis Laplacian weight saliency weight, fuse computed kernels three channels improve quality experimental results show that our algorithm has