THRESHOLDING FOR IMAGE SEGMENTATION USING 2D-HISTOGRAM AND SPECTRAL CLUSTERING

Traditional 2D thresholding methods suppose that the sum of probabilities of main-diagonal distinct in 2D histogram is approximately one, and neglect the sum of probabilities of counter-diagonal distinct in 2D histogram. Therefore, the assumption mentioned above is inadequately reasonable; at the same time, those methods become very time-consuming when extended to a multi-level threshold problem due to the fact that a large number of iterations are required for computing the cumulative probability and the mean of a class. To overcome the shortcomings mentioned above, a new 2D thresholding algorithm using 2D histogram and spectral clustering algorithm is presented. Spectral clustering algorithm can correctly recognize the arbitrary shaped clusters. Experiments based on infrared images and real images demonstrate the advantages of the proposed algorithm.