Wavelet transform approach to segment thermal image

A multi-threshold approach to segment thermal image is given based on wavelet transform. The gray-level histogram of original image is obtained. In order to reduce the effect of noise, the gray-level histogram is smoothed by Bezier curve and called Bezier histogram. The wavelet transform with dimensional stability is used to the curvature curve of Bezier histogram. Positions of peak values of curvature curve in wavelet domain are adjusted from fine to coarse at all scales. The gray level values, which are located in adjusted peak values at all scales, are considered as segmentation thresholds. The gray level values of valley between two peaks are considered as quantity gray levels. Optimal segmentation scale is obtained by a cost criterion.