Object Detection Using Color Dissimilarity Based Segmentation Method

Image segmentation is a very important process in object detection. The segmentation process becomes a critical determinant of the success of object detection. Various methods have been developed for this image segmentation process, but there is no general solution that can be applied. Associated with object detection, the image segmentation method based on color dissimilarity turns out to be able to give good results. This segmentation method divides the image based on the dissimilarity of the values of R, G and B on the adjacent image pixels. Color is considered different if the difference in the values of R, G and B of this pixel pair produces a Delta E value whose value exceeds the threshold that is able to distinguish eyes. If this comparison shows the color dissimilarity detected, the pixel is changed to white, and the resulting image is then converted into a black-and-white image. From the resulting segmentation space, the bounding box is then made. Based on this bounding box, then the object can be detected properly. The results of object detection shown by this method are very good. The weakness of this model is not being able to detect objects that overlap one another.