一种融合色彩和亮度信息补偿的图像预处理方法

An Image Preprocessing Method Combining Color and Brightness Information Compensation

双目视觉算法中,需要对获得的图像进行预处理。常用的处理方法是进行去噪声和直方图均衡,虽然能改善相机畸变和光照异常带来的影响,但忽略了色彩本身对特征点的影响。该方法首先结合人脸肤色特点采用参考白的方法对亮度进行了补偿,而后经过同态滤波,将颜色变化转化为亮度变化,增强了图像细节,最后采用尺度不变特征变换(Oriented FAST and Rotated BRIEF,ORB)算法对实验的若干幅图像提取特征点。实验分析表明,该方法减少了冗余特征点,使分布均匀化,避免了特征匹配中出现过多的误匹配对,降低了匹配难度。

In the binocular vision algorithm,it is necessary to preprocess the obtained image,and the commonly used processing methods are noise removal and histogram equalization.Although it can also improve the impact of camera distortion and abnormal illumination,it ignores the impact of color itself on feature points.The method first uses the reference white method to compensate the brightness based on the characteristics of the skin color of the face,and then after homomorphic filtering,the color is converted into brightness changes,which enhances the image details,and finally,the ORB algorithm is used to extract feature points from several experimental images.Experimental analysis indicates that this method can decrease redundant feature points,make the distribution uniform,and avoid too many false matching pairs in feature matching,and reduce the matching difficulty.