Fast extraction of linear segmentation characteristic based on gray scale projection multiple processing

The fast extraction of segmentation characteristic from images constructed by complex objects and contaminated by noise is a difficult problem in image process domain. A fast extraction algorithm of linear segmentation characteristic based on multiple processing to gray scale projection is proposed. The original gray scale projection of image is mapped with rectangle window to obtain statistics distribution curve. Classical first difference calculator is used to extract the extreme point on the curve. The coordinate of linear segmentation characteristic formed by expected region edges is obtained based on the extreme point gradient criteria. Demonstrated by application, the algorithm is accurate and effective for optical noisy images.