Real Time Colour Image Segmentation with Non-Symmetric Gaussian Membership Functions
Omid Sojodishijani, Vahid Rostami, Abd Rahman Ramli · 2008
Segmentation and also partitioning image into disjoint subsets are one of the most important stages in many computer vision applications. In this paper the new algorithm for dividing colour image into homogeneous regions is proposed. In this algorithm we use histogram estimation in 3 channels - RED, GREEN and BLUE - to create adaptive membership functions that update dynamically when the entire image is scanned. These triple membership functions characterise each region in image. For producing the proper estimation of histograms, the non-symmetric gaussian function (NSGF) is proposed. To achieve real time implementation, the algorithm scans the image in raster fashion. Finally, the performance of algorithm is shown in the result of applying this algorithm to natural and synthetic noisy images.