A video-rate color image segmentation using adaptive and statistical membership function

The color image segmentation is a critical task in many computer vision applications. The function of segmentation is to identify homogeneous regions in an image, based on properties such as intensity, color and texture. Typically, the image segmentation algorithms in video processing system require very high computation power, so it is desirable to develop algorithms for implementation as a real-time system. This paper proposes a novel image segmentation algorithm and its real-time hardware architecture which is capable of dealing with regions color information. In this algorithm, statistical information of regions is used to create fuzzy membership functions in color model components. These membership functions characterize each segment in an image, which are updated dynamically when the image is being scanned. The histogram of color components are estimated by non-symmetric Gaussian function (NSGF). To overcome the video-rate limitation, the image is scanned in the raster fashion. Moreover, the hardware architecture of algorithm on FPGA is reported in this paper. Finally, the results of algorithm are analyzed by quantitative performance analyzers.

Text 12896.pdf Restricted to Repository staff only Download (932kB) Official URL or Download Paper: http://www.academicjournals.org/journal/SRE/articl...

Actions (login required)

View Item