Abstract

For the Image threshold segmentation, local contrast information is not sufficient used for threshold, the border is a single pixel wide, but is deviation from the true boundary point. Gradient segmentation does not take full advantage of global information, which can get the real boundary points, and affected by noise with not complete single-pixel contour. By analyzing the threshold segmentation and gradient segmentation, fused with which a segmentation algorithm is proposed. The histogram threshold is studied in three ways: interactive, iterative and the Otsu method, and the wide valley impact on image segmentation is analyzed, the algorithm of selecting the best threshold is designed. The optimal threshold is researched on the gradient mean of the same gray value, which integrates the local gradient and global threshold. Experimental results show that the algorithm not only considers the adjacent consistency of the image information, but also taking into account the local contrast, and improve the effect of simple threshold, which is an effective and practical image segmentation algorithm.

Paper Details