Color Image Segmentation in RGB Color Space Based on Color Saliency.
Chen Zhang,Wenzhu Yang,Zhaohai Liu,Daoliang Li,Yingyi Chen,Zhenbo Li
DOI: https://doi.org/10.1007/978-3-642-54344-9_41
2013-01-01
Abstract:It is difficult to separate the foreign fiber objects in a captured color image from the background accurately. To solve this problem, this paper presents a new approach for color image segmentation in RGB color space based on color saliency. Firstly, the captured RGB color image was separated to R, G and B color channels, and the color information for each channel was calculated. Secondly, the R saliency for each pixel in the R channel, the G saliency for each pixel in the G channel, and the B saliency for each pixel in the B channel was calculated respectively. Then comprehensive saliency map was obtained by the weighted R, G and B saliency. The weights for the R, G and B saliency were determined by the corresponding color information of each color channel. At last, the foreign fiber targets were separated out from the comprehensive saliency map using a threshold method. The results indicate that the proposed method can segment out the foreign fiber objects from the color image accurately.