为提高图像的可视化效果, 增强图像的局部细节信息, 提出一种结合引导滤波和非线性二阶特征的色调映射方法, 首先提取输入图像的亮度信息, 利用引导滤波器对亮度图像进行多尺度分解, 得到基本层和细节层图像, 其次, 通过引导滤波方法构造细节层的权重图, 对基本层利用Hessian矩阵构造一种非线性二阶特征, 再通过引导滤波方法构造基本层的权重图, 最后, 根据权重图对分解后的基本层和细节层图像实现亮度图像的重构, 然后恢复亮度图像的色彩信息, 获得最终的结果图像。实验结果表明, 该方法较为完整的保留源图像的局部细节信息, 具有良好的视觉效果。本文将Hessian矩阵用于提取基本层图像的高频信息, 可以更好地突出图像的边缘信息, 丰富图像的细节特征。与对比算法的客观指标相比, 该方法的质量分数提高了11.28%, 结构保真度提高了10.82%, 自然相似性提高了186.46%。

To improve the visualization effect and enhance the local detail information of an image, a tone-mapping method that combines guided filtering and nonlinear second-order features was proposed. First, the luminance information of an input image was extracted and the luminance image was decomposed by a multi-scale using a guide filter to obtain a base layer and a detail layer image. Second, a weighting map of the detail layer was constructed by the guided filtering method, a nonlinear second-order feature was constructed using a Hessian matrix for the basic layer, and a weight map of the basic layer was constructed using the guided filtering method. Finally, the basic graph was decomposed according to the weight map. The layer and detail layer images implement reconstruction of the luminance image and then restore the color information of the luminance image to obtain the final resulting image. Experimental results show that the method preserves the local detail information of the source image and has a good visual effect. In this study, the Hessian matrix is used to extract the high-frequency information of the base-layer image, which can more effectively highlight the edge information and enrich the detail features of the image. Compared with the objective index of the comparison algorithm, the quality score of the method, the structural fidelity, and the natural similarity are increased by 11.28%, 10.82%, and 186.46%, respectively.