Color Correction Algorithm Research Of Four-Band Image Based On Color Constancy

Color constancy is a characteristic of the human color perception system. The eyes can identify the color of object under the different illuminations automatically. But this is only for human eye, camera is affected by light easily, So the color of the image has changed. Color correction is an essential part of image processing.The research object of this paper is four band image which is taken by visible light with infrared light (850nm). The method in this paper is using algorithm to eliminate the influence of infrared light instead of filters. This algorithm contains two steps, First, because R channel and B channel are influenced by infrared light badly, so a preliminary correction on R and B channels in this paper, then get a color cast image, But the color of the image has a clear distinction at this time. Second, Frankle-McCann Retinex based on color constancy is proposed to correct the Color cast images. In this paper, the Frankle-McCann Retinex algorithm compared with the traditional Frankle-McCann Retinex algorithm has two improvements, one is to add a distance-weighting factor with Gauss when the pixels are compared, the other is to add a linear stretch with the mean and the standard deviation at the end of algorithm. Experimental results show that this method has better correction effect, which lays foundation for subsequent research of four band image.