Estimation of Low Dynamic Range Images from a Single Image for High Dynamic Range Imaging Based on Exposure Look-up Table

摘要

HDR(high dynamic range) imaging supports wider dynamic range than normal images captured from general still camera. However, this technique usually needs several shots to obtain LDR(low dynamic range) images as merging purpose, causing artifacts such as ghosting. This paper suggests a method to estimated necessary LDR images from a single image using exposure LUT(look-up table) constructed by simple measurement. We first construct exposure LUT for each RGB channel. Then, from an image captured by auto mode in camera, LDR images which have each target average luminance are be estimated using constructed LUT. All procedures are performed with RGGB Bayer image. Also, saturated area is compensated by considering channel dependency. In experimental comparison, high PSNR values are obtained between estimated and captured images.