受环境因素的影响，金属材料表面会呈现出不同的腐蚀特征，如形貌、纹理、颜色等，通过对这些特征的分析，可以判断出材料的腐蚀程度以及腐蚀类型。然而，由于受到天气、光照和成像设备等外界因素的影响，使得这些腐蚀特征的呈现变模糊，直接影响了腐蚀图像中腐蚀细节信息的展现与传达。因此，为了更准确地检测与识别腐蚀形貌，对采集到的原始低照度腐蚀图像有针对性的增强，具有十分重要的意义。

针对低照度腐蚀图像整体偏暗、对比度低、腐蚀细节模糊等缺陷，本文提出了一种改进的同态滤波算法与多尺度融合的腐蚀图像增强算法。该算法首先对V分量采用改进后的单参数分块同态滤波得到亮度增强图像；其次利用多尺度加权引导滤波将原始腐蚀图像分为基础图像和细节图像后加权融合，获得细节对比度增强图像；最后将亮度增强图像与细节对比度增强图像多尺度融合得到最终增强腐蚀图像。实验结果表明，与原始腐蚀图像相比，增强后图像的标准差平均提升了约18%，平均梯度与信息熵分别平均提升了约43%和12%。改进的算法能够有效提升腐蚀图像亮度与对比度，增强后的腐蚀图像可以展现更多信息。

针对低照度腐蚀图像的降噪性能差，以及腐蚀图像增强后易产生色彩失真等问题，提出了改进小波变换与多尺度Retinex融合的腐蚀图像增强算法。该算法首先将V分量的低照度腐蚀图像通过小波变换分解成高频分量和低频分量；采用改进的小波阈值函数去除高频分量的噪声，采用基于增强二维gamma函数改进的多尺度Retinex算法校正低频分量的光照信息；然后将小波重构后的图像同样与细节对比度增强图像多尺度融合来增强腐蚀区域对比度，从而得到最终增强图像。实验结果表明，与原始腐蚀图像相比，增强后图像的标准差平均提升了约32%，平均梯度与信息熵分别平均提升了约49%和27%，峰值信噪比高于其他算法。改进后的融合算法抗噪声能力较强，并且腐蚀图像的色彩恢复能力较强，显著提升了腐蚀图像整体质量。

Due to the influence of environmental factors, the surface of metal materials will show different corrosion characteristics, such as morphology, texture, color, etc., through the analysis of these characteristics, the degree of corrosion of the material and the type of corrosion can be judged. However, due to the influence of external factors such as weather, lighting and imaging equipment, the presentation of these corrosion characteristics becomes blurred, which directly affects the display and communication of corrosion details in the corrosion image. Therefore, in order to detect and identify corrosion morphology more accurately, it is of great significance to enhance the original low-illumination corrosion image collected.

Aiming at the overall darkness, low contrast, and blurred corrosion details of low-illumination corrosion images, an improved homomorphic filtering algorithm and a multi-scale fusion corrosion image enhancement algorithm are proposed. Firstly, the improved single-parameter block homomorphic filtering is used for the V component to obtain the brightness enhancement image. Secondly, multi-scale weighted guided filtering is used to divide the original corrosion image into basic image and detail image and then weighted fusion to obtain detailed contrast enhancement image. Finally, the brightness-enhancing image and the detail-enhancing contrast-enhancing image are fused at multiple scales to obtain the final enhanced corrosion image. The experimental results show that compared with the original corrosion image, the standard deviation of the enhanced image is increased by about 18% on average, and the average gradient and information entropy are increased by about 43% and 12%, respectively. The improved algorithm can effectively improve the brightness and contrast of the corrosion image, and the enhanced corrosion image can show more information.

Aiming at the poor noise reduction performance of corroded images and the easy color distortion after corrosion image enhancement, a corrosion image enhancement algorithm is proposed to improve the fusion of wavelet transform and multi-scale Retinex. Firstly, the low-illumination corrosion image of V component is decomposed into high-frequency components and low-frequency components by wavelet transform. The improved wavelet threshold function is used to remove the noise of the high-frequency components, and the multi-scale Retinex algorithm based on the enhanced two-dimensional gamma function is used to correct the illumination information of the low-frequency components. Then, the wavelet reconstructed image is also fused with the detail contrast enhancement image at multiple scales to enhance the contrast of the corroded area, so as to obtain the final enhanced image. The experimental results show that compared with the original corrosion image, the standard deviation of the enhanced image is increased by about 32% on average, the average gradient and information entropy are increased by about 49% and 27%, respectively, and the peak signal-to-noise ratio is higher than that of other algorithms. The improved fusion algorithm has strong noise resistance and strong color recovery ability of corroded images, which significantly improves the overall quality of corroded images.