高动态范围(HDR)图像指的是具有更高的能被人眼识别的亮度动态范围的图像,它能够更加全面地展现场景的细节信息。针对单幅低动态范围(LDR)图像,提出一种基于多层伽马变换融合的HDR图像生成方法。对LDR图像统计特性进行分析,将其分成4个亮度等级区域,每个区域自适应生成伽马变换参数;将得到的4个伽马变换参数依次作用在原图像上,得到强调不同区域细节信息的4幅图像;将4幅伽马变换后的图像融合生成HDR图像。展示利用本文方法生成的HDR图像色调映射结果,并与基于多幅伪曝光图像融合生成HDR图像的算法进行对比,结果表明,本文方法所生成的图像具有更高的信息熵,且算法运行时间更短。

High dynamic range (HDR) images are the images with HDR of lightness that can be recognized by human eyes, which can show more details about scenes. Aimed at one-exposure low dynamic range (LDR) images, we propose an HDR image generation method by fusing multi-level gamma-transformed images. The LDR image is firstly divided into four regions according to its histogram distribution. The Gamma parameter is calculated in each region based on the lightness level. Then, four images with richer details of different regions are generated by multi-level Gamma transform. Finally, they are fused into one HDR image with higher dynamic range and richer details. The tone-mapped HDR images generated by proposed method are compared with those by fusing multiple pseudo-exposure images. The results show that the proposed method has higher entropy and shorter running time.