从人眼亮度感知特性出发, 提出了一种人眼感知驱动的成分分解色调映射算法。通过非线性全局亮度校正函数调整整体亮度; 通过引导滤波将图像分解为基本层与细节层, 并在基本层结合人眼感知特性进行动态范围压缩; 对Stevens效应下的细节层进行效果增强, 使映射后的图像细节更加丰富。对比传统算法, 不同场景的实验结果显示, 该算法的主观质量与客观指标均有所提高, 在主观质量上能体现HDR图像高亮度光源照射的场景特点, 细节表现更加清晰, 图像层次感有所提高; 在客观指标上, 信息熵平均提高0.2291, TMQI值平均提高0.0889。

A perception-inspired tone mapping(TM) algorithm based on components decomposition is proposed. The nonlinear global brightness correction function is used to adjust the overall brightness of the HDR image. The basic layers and the detailed layers are separated via the guided filter, and the dynamic range of basic layers is compressed with human brightness perception characteristics. The local details are enhanced in the detailed layers under the effect of Stevens to obtain the more abundant images. The experimental results show that compared with the traditional TM approach, of the algorithm can reflect the more scene features of the image under the high brightness light source and gain more details of the image in the subjective perception. Meanwhile in the objective evaluation, the algorithm enhances the average information entropy 0.2291 and the average TMQI 0.0889.