Deblurring of Document Images Based on Sparse Representations Enhanced by Non-local Means

Abstract

Blur is one of the most difficult distortions in camera captured documents. It degrades the visual quality of an image, and makes it difficult to read whether by a human or OCR systems. This paper presents a novel non-blind deblurring method that combines the well known effective techniques of sparse representations and non-local image similarity. The presented problem formulation enables the use of standard sparse coding methods for solving sparse coding-based deblurring when enhanced by a non-local means prior. The method has been tested on both synthetic and real document images degraded with a variety of blur kernels. The resulting deblurred images have high quality in terms of both signal-to-noise ratio and OCR accuracy.

Blur2Sharp: A GAN-Based Model for Document Image Deblurring

“…Pan et al [5] proposed another approach for text deblurring that makes profit of L0 regularized intensity and gradient priors. Nayef et al [6] suggested a method for document image deblurring that uses sparse representations improved by non local means. Zhang et al [7] used a gradient histogram preservation strategy for document image deblurring.…”