Automatic Image Segmentation Using Marker Controlled Watershed And Overlap Ratio Based Region Merging.

In this paper, automatic image segmentation system is proposed to obtain accurate and meaningful segmented regions in an image with less over-segmentation. It includes three main approaches: preprocessing, segmentation and post processing steps. In the preprocessing step, the modified 7x7Laplacian of Gaussian (LoG) edge filter is implemented to compute the accurate approximation of gradient magnitudes. In segmentation step, marker controlled watershed method (MCWS) is applied on image gradient magnitude. Finally, the over-segmented regions are merged by using histogram similarity to merge the homogeneous regions. This system intends to produce the correct, useful and meaningful segmented results for medical analyzing tasks, objects detection and recognition in an image. It is tested on two different kinds of datasets: medical images and color natural image dataset. This system has also achieved accuracy 93.01% for MRI brain images, 76.72% for color natural images.