单幅图像的高动态范围图像生成方法

摘要

In order to convert the existing low dynamic range (LDR) image into the high dynamic range (HDR) image, a method of generating HDR image from a single LDR image is proposed in this paper. Firstly, the input LDR image is decomposed into luminance component and chrominance component based on human visual sys-tem, then, constructing the inverse tone mapping operators of this luminance and obtaining its inverse tone mapped image by pixel-wise mapping. Secondly, defining a threshold to sample on the high light areas of the lu-minance to get a light-source sampled image, and extending the obtained light-source sampled image by Gaussian filter and erosion which can simulate the light attenuation. Thirdly, merging the inverse tone mapped image with the extended high-light-area image by constructing the piece-wise function to get the final luminance. Finally, combining this luminance with the chrominance to generate the HDR image. A plenty of experiments show that the proposed method can get HDR image from a single LDR image with good robustness, which performs better and faster than previous methods.%为了将现有低动态范围(LDR)图像转换为高动态范围(HDR)图像，提出一种将单幅 LDR 图像转换成对应HDR 图像的方法。该方法基于人类视觉系统(HVS)模型，首先分离出 LDR 图像的亮度分量和色度分量，对亮度分量构造其反色调映射函数，通过逐点映射得到亮度分量的反色调映射图像；然后对亮度分量进行光源采样处理，并采用高斯滤波和腐蚀操作模拟光线衰减效应，对光源采样图像进行高光区域的扩展；再构造分段函数对反色调映射图像和高光区域扩展图像进行融合，得到最终的亮度分量；最后合并该亮度分量和色度分量得到最终的 HDR 图像。实验结果表明，文中方法能通过单幅LDR图像得到HDR图像，处理效果较好，运行效率高，具有较好的鲁棒性。