Over-exposure image correction with automatic texture synthesis

A novel over-exposure image correction algorithm combining synthesized texture information in over-exposed regions is presented. In this algorithm, intensity image is decomposed into structure and detail layers by weighted least squares method. Modified exemplar-based texture synthesis method is applied to the detail layer in order to synthesize texture details in over-exposed regions. The synthesized texture is combined with the original texture and fed to the lightness and color correction modules to get the final result. An efficient diffusion-like color correction method is presented. Qualitative and quantitative experimental results show that the proposed method produces vivid details which are absent in previous methods without human assistance in over-exposed regions. This improves the overall appearance of the recovered image.

I. Introduction

The natural world contains a wide range of light intensities which can not be captured by most of the consumer grade photographic sensors. For example, the relative brightness of an outdoor scene in a sunny day is about 400,000. Even for the normal indoor environment, the relative brightness will still reach to about 1,500 [1]. However, the dynamic range of common digital camera is only 1000:1. As a result, overexposure is inevitable in daily-life photography.