Retinex-based color correction for displaying High Dynamic Range images

Until now, most of methods for High Dynamic Range (HDR) images focus on tone mapping. They can compress large range luminance values so that they can be displayed on ordinary devices. However, the lighting condition in which the images are taken is not considered. As a result, the mapped images may have color cast. Therefore, to get an image that is similar to direct observation of a scene for human eyes, tone mapping alone for a HDR image is not sufficient. In order to solve the problem, we propose an Orientation-based Retinex (O_Retinex) method serving as color correction for HDR images. On the other hand, local contrast loss may be caused by most tone mapping methods when the dynamic range of a HDR image is quite high. In our paper, a contrast enhancement technique is proposed to solve the problem. Furthermore, the color correction and the tone mapping are processed in parallel for our method, which is the main difference from other methods that serve the tone mapping as a preprocessing. Because the proposed method uses the parallel processing strategy, it can adopt most of tone mapping methods. The only constraint on them is that the ratios among three color channels must be preserved. Experiments show that the proposed method can efficiently reduce color cast induced by lighting condition for HDR images.