Text Image Deblurring Using Text-Specific Properties

Abstract

State-of-the-art blind image deconvolution approaches have difficulties when dealing with text images, since they rely on natural image statistics which do not respect the special properties of text images. On the other hand, previous document image restoring systems and the recently proposed black-andwhite document image deblurring method [1] are limited, and cannot handle large motion blurs and complex background. We propose a novel text image deblurring method which takes into account the specific properties of text images. Our method extends the commonly used optimization framework for image deblurring to allow domain-specific properties to be incorporated in the optimization process. Experimental results show that our method can generate higher quality deblurring results on text images than previous approaches.

Deblurring Text Images via L0-Regularized Intensity and Gradient Prior

“…where σ is the weight defined in (2). When β and µ are close to ∞, the solution of (6) approaches that of (4).…”

“…However, this method is only applied to two-tone images and is less effective for text images with complex backgrounds. Cho et al [2] develop a method to incorporate text-specific properties (i.e., sharp contrast between text and background, uniform gradient within text, and background gradient following natural image statistics) for deblurring. While this algorithm achieves the state-of-the-art deblurring results, the kernel estimation process is complicated and the performance depends largely on whether the stroke width transform (SWT) [5] separates an image into text and nontext regions well or not.…”

“…If the characters in a text image are small and connected, it is unlikely to perform well. Figure 1 shows one example, where blurred characters are connected due to large camera motion, and the deblurred result from the state-of-the-art algorithm [2].…”

“…We present analysis on the relationship with other methods based on salient edges, and show that the proposed algorithm generates reliable intermediate results for kernel estimation without any ad-hoc selection processes. Compared to the state-of-the-art methods [1,2], the proposed algorithm is simple and easy to implement as it requires no additional operations (e.g., adaptive segmentation [1], smoothing intermediate latent images, or SWT [2]). In the latent image restoration step, we present a simple method to deal with artifacts.…”

A MAP-Estimation Framework for Blind Deblurring Using High-Level Edge Priors

“…In blind deconvolution, the idea of exploiting domain-specific properties had been previously used for the deblurring of text images [3]. More generally, using scene-specific prior knowledge for improving the performance of computer vision algorithms is an idea that has already been applied with great success in other contexts in the past.…”