基于梯度稀疏和多尺度变分约束的图像增强算法

针对低照度条件下采集到的图像存在亮度偏低、细节模糊等问题，通过分析传统Retinex理论在增强图像过程中的缺陷，提出了一种基于梯度稀疏和多尺度变分约束的图像增强算法。该算法首先将输入图像由RGB空间转换到HSV空间，提取亮度分量，实现三个通道的解耦合。然后根据零范数的梯度全局显著特性，定义了一个新的相对全变分正则项。接着在HSV空间下惩罚亮度分量，构建一种具有梯度稀疏的变分模型对亮度通道进行约束，并通过将控制因子扩张为多个尺度，形成多尺度变分约束，提升照度估计的准确度，使之更加符合光照分布特性。根据Retinex理论进行映射，获取亮度通道对应的反射图像。进而利用亮度通道不同尺度下的约束所对应的不同照度结果，分别提取图像的粗略细节、中等细节和精细细节，通过多尺度细节加权，对反射图像进行细节增强。最后，对照度图像进行伽马校正，与经细节提升后的反射图像重组并进行颜色空间转换得到输出的增强图像。通过实验对比表明，所提算法的增强图像有着更高的色彩丰富度和更低的色差水平，能够保持图像的自然度，提升图像的视觉效果。在均值、平均梯度和信息熵的表现上，相比原图均有大幅度提升，与现有的先进算法相比，平均定量指标在不同类型低照度图像的增强图像上均产生了较优的效果，且有着较快的运算效率。

Image Enhancement Algorithm Based on Gradient Sparsity and Multi-scale Variational Constraint

Aiming at the problems of low brightness and blurred details in the images collected under low illumination, an image enhancement algorithm based on gradient sparsity and multi-scale variational constraint was proposed by analyzing the defects of traditional Retinex theory in the process of image enhancement. Firstly, the input image was transformed from RGB space to HSV space, and the luminance component was extracted to realize the decoupling of three channels. Then, according to the gradient global significance of zero norm, a new relative total variation regular term was defined. After that, the luminance component was punished in HSV space, and a variational model with gradient sparsity was constructed to constrain bright channel. By expanding the control factors to multiple scales, a multi-scale variational constraint was formed, which improves the accuracy of illumination estimation and makes it more in line with the illumination distribution characteristics. According to Retinex theory, the reflection map corresponding to brightness channel was obtained. Then, the rough details, medium details and fine details of the image were extracted by using different illumination results corresponding to constraints in different scales of bright channel, and the details of the reflection map was enhanced by multi-scale detail weighting. Finally, gamma correction was carried out on the illumination map, the image was recombined with reflection map after detail enhancement, and color space conversion was carried out to obtain the output enhanced image. Experimental comparison shows that the enhanced images of the proposed algorithm has higher color colorfulness index and lower color difference level, and can keep the naturalness and improve the visual effect of images. Compared with the original image, the performance of mean value, average gradient and information entropy has been greatly improved. Compared with the existing advanced algorithms, the average quantitative index has a better effect on the enhanced images of different types of low illumination images, and has faster operation efficiency.