Text Image Deconvolution Using Re-blurry Process And $L_{0}$ Norm
Xiaoyuan Yu, Bo Li, Jinwei Yu, Wei Chau Xie · 2020
For single image blind deblurring issue, we need to estimate the kernel and the latent image from an observed blurry image. It is well known that the statistical results of clear text image with re-blurring process is very different with that of blurry text image. Furthermore, according to the prosperity of text image, the L0 norm has been proofed effectively in text image deblurring. Therefore, inspired by observing distinct properties of text image with or without re-blurring process, we propose a new deblurring method to process text image by using the re-blurry process and L0 norm. First, we employ the proposed regularization terms to the image deblurring framework. Then, we propose an effectively optimized method to estimate kernel and latent image. Finally, comparing with the state-of-the-art methods are evaluated quantitatively and qualitatively. In conclusion, the proposed method performs favorably against the state-of-the-art methods on processing blurry text images.