COMBINATIONAL EDGE DETECTION USING MULTIPLE COLOR CHANNELS AND GRABCUT

Identification of edges is very important in feature extraction and pattern recognition. An edge of an image detected by converting it from RGB format to a Grayscale image and can sometimes be inefficient and inaccurate. This inefficiency is caused due to various color differences that get erased or rewritten during the process of grayscale conversion. Edge locations in a colored image are derived by analysing the variations in the multiple color channels and merging the gradients in these channels to compute a single edge snapping vector field, which is derived using the Euclidean distances between two distinct pixels in an image. This lets to retain the various multidimensional characterizations of color channels of an image as the color differences can be closely calculated using human perceptions. Hence this method is proved to provide more accuracy in edge detection. Furthermore, the results are improved when combined with a finer edge detection method GrabCut, which allows the user to detect the edges in an image using reduced iterations. The proposed paper uses a combinational approach to efficiently acquire edges in the image by enhancing the color properties of an image and then using the GrabCut method to retrieve the edges present in an image.