Natural images contour segmentation / Khairul Adilah Ahmad, Sharifah Lailee Syed Abdullah and Mahmod Othman

This paper, a combination of edge detection and contour based segmentation approach for object contour delineation is proposed. The proposed approach employs a new methodology for segmenting the fruit contour from the indoor and outdoor natural images more effectively. The overall process is carried out in five steps. The first step is to pre-process the image in order to convert the colour image to grayscale image. Second step is the adoption of Laplacian of Gaussian edge detection and a new corner template detection algorithm for adjustment of the pixels along the edge map in the interpolation process. Third step is the reconstruction process by implementing two morphology operators with embedded of inversion condition and dynamic threshold to preserve and reconstruct object contour. Fifth step is ground mask process in which the outputs of the inference obtained for each pixel is combined to a final segmented output, which provides a segmented foreground against the black background. This proposed algorithm is tested over 150 indoor and 40 outdoor fruit images in order to analyse its efficiency. From the experimental results, it has been observed that the proposed segmentation approach provides better segmentation accuracy of 100 % in segmenting indoor and outdoor natural images. This algorithm also present a fully automatic model based system for segmenting fruit images of the natural environment.