摘要
非正面人脸图像比正面人脸图像更容易产生因角度原因导致光照不均匀进而肤色出现色彩偏差的问题,为了应对这种问题,详细分析了灰色世界法、完美反射法、动态阈值法和参考白算法,并在非正面人脸库进行实验。经过研究发现,可以根据平均亮度来选择光照补偿算法。实验结果表明,在平均亮度小于100时,采用参考白算法,平均亮度大于150时,采用完美反射法。在检测效果上比单一的光照补偿算法更优。
Non-frontal face images are more likely to cause uneven illumination and color deviation due to angle problems than frontal face images.In order to deal with this problem,the gray world method,perfect reflection method,dynamic threshold method and reference white are analyzed in detail in this paper.These four illumination compensation algorithms are experimented with non-frontal face database.After research,it is found that the illumination compensation algorithm can be selected according to the average brightness.Experimental results show that when the average brightness is less than 100,the reference white algorithm is used,and when the average brightness is greater than 150,the perfect reflection method is used,which is better than a single illumination compensation algorithm in terms of detection speed and effect.
出处
《工业控制计算机》
2021年第1期53-54,56,共3页
Industrial Control Computer
关键词
非正面人脸
肤色检测
完美反射法
参考白算法
平均亮度
non-frontal face
skin color detection
perfect reflection method
reference white algorithm
average brightness
