Nonlinear diffusion equation with selective source for binarization of degraded document images

Binarization has always been a very challenging task for degraded document images be-cause of the variety and complexity of degradations. This paper proposed a nonlinear diffu-sion equation with selective source for restoration of degraded document images, followed by a binary projection for binarization. The source is composed of two parts: one is used for the restoration of contaminated background, and another is responsible for the fidelity of texts. The evolution equation is numerically solved by the simplest finite differencing. The proposed method is tested on publicly available datasets (DIBCO (Document Image Bi-narization Competition) 2009-2014, 2016) and is compared with five models based on par-tial differential equation (PDE). The experimental results show that the proposed method is effective for degraded document images and has achieved generally the best performance of binarization compared to the other five PDE methods. (c) 2021 Elsevier Inc. All rights reserved.