Patch-based over-exposure correction in image

Most of conventional over-exposure correction methods fail to reconstruct the detail information of the overexposed region (OER) in an image. This paper presents a novel method which can effectively reconstruct the texture as well as the luminance and the color of the OER without user interaction. The proposed method is performed based on a patch-based region filling method. In a patch including the over-exposed pixels, new luminance values are progressively estimated according to the luminance variation in the reference patch (RP) which is selected in the well-exposed region. For the color correction of the OER, the color values of the RP are simply copied. Once the OER is completely filled, the resultant image is refined using the Gaussian-based filters. Experimental results demonstrate that the proposed method provides more preferable results than conventional one.