基于 Retinex 的人脸光补偿算法研究
A Research on Face Light Compensation Based on the Retinex Algorithm
-
摘要: 本文研究了基于图像处理技术的光照补偿方法, 针对直方图均衡化对, 图像背景太亮或太暗有很好的处理效果, 但对细节处理效果很差的问题结合 Retinex 对于恢复图像细节效果强的优点, 提出了基于小波变换的融合方法, 将图像分别由直方图均衡化处理和 Retinex 处理后的两幅图像进行融合, 并通过 Yale B 人脸库进行相应的实验。实验结果对比表明, 采用融合方法可以弱化光照对图像的影响, 同时对图像的细节处理效果明显, 增强了图像的特征信息, 提高了图像的清晰度。Abstract: Illumination processing method based on image processing technology is studied. Histogram equalization is effective for treating too bright or too dark background images but is very poor in detail treatment. Retinex is effective in restoring image details. Combining histogram equalization and Retinex, a fusion method is proposed based on wavelet transform. The corresponding experiments were conducted on Yale B face database. The comparison of the experiments showed that the fusion method can weaken the impact of the light on the image, properly preserve the details with better effect of processing, enhance the image feature information and improve the clarity of the image.