Segmentation of wood surface knots and wormholes based on an improved LBF Model

Abstract

To make wood surface defect segmentation faster and more effective, research was conducted to put forward an improved LBF （local binary fitting） Model with image segmentation based on the Chan-Vese （CV） Model. The improved LBF Model added a new level set formulation with a linear regularization term, and at the same time formed a Gaussian function as the kernel function with two local values for fitting energy. Results showed that the improved algorithm could overcome the segmentation shortcomings in the LBF Model. Also, the segmentation process was not sensitive to the size or the position of the initial contour. However, the anti noise of the algorithm was enhanced, and the image could be segmented in non-uniform gray. The experiment showed that the algorithm completely extracted the wood surface defect images with single and multi objectives, and level set evolution corresponding to the defect image could be obtained. ［Ch, 21 fig, 1 tab. 15 ref.］