融合非物理模型的改进AM-RetinexNet图像增强算法

Improved AM-RetinexNet Image Enhancement Algorithm Incorporating Nonphysical Models

针对地下空间低照度图像色彩偏暗、亮度低且分布不均、增强后图像色偏和噪声高等问题,研究提出了融合非物理模型的改进AM-RetinexNet图像增强算法。该算法将RGB图像转换成HSV分量,利用HSV空间相互独立性实现图像亮度增强和色彩增强处理,其中S分量利用V分量提取的相关信息进行自适应调整,V分量采用融合直方均衡化与注意力机制优化的RetinexNet进行照度分量增强处理;将处理后HSV分量转化成RGB图像,并对图像进行自适应色彩恢复,得到照度增强图像。对比实验表明,在图像的细节处理、亮度整体增强处理、图像降噪和色彩视觉修正等方面该方法表现较好,测试指标中平均互信息(MI)、标准差(STD)、结构相似性(SSIM)、平均梯度(AG)、空间频率(SF)和峰值信噪比(PSNR)最佳,均值分别可达到6.18,70.62,0.56,13.29,36.53,39.22。

To solve the problems such as dark color, low brightness and uneven distribution, color deviation after image enhancement and high noise of low illumination image in underground space, an improved AM-RetinexNet image enhancement algorithm combining non-physical models is proposed. The algorithm converts RGB image into HSV component, and makes use of the independence of HSV space to achieve image brightness enhancement and color enhancement. Among them, S component uses the information extracted from V component for adaptive adjustment, and V component uses RetinexNet optimized by fusion of straight-square equalization and attention mechanism for illumination enhancement. Finally, the processed HSV component is transformed into RGB image, and the image is restored by adaptive color to obtain illumination enhanced image. Experimental results show that the method performs well in image detail processing, luminance enhancement, image denoising and color visual correction. The average MI, STD, SSIM, AG, SF, PSNR perform best among all metrics, with mean values respectively reaching 6.18, 70.62, 0.56, 13.29, 36.53, 39.22.