Fast multilevel thresholding for image segmentation through a multiphase level set method

For the image segmentation by the histogram bilevel thresholding, several methods have been proposed. However, they are computationally time consuming and their effectiveness is reduced when applied to a complex image and when the number of the different regions composing this image is high. In this paper, a fast and efficient method for segmenting complex images is proposed. This method is based on the determination of the number and the values of the thresholds required for the segmentation by introducing a new multilevel thresholding technique using a multiphase level set technique. First, the gray-level histogram of the image is approximated by a weighted sum of Heaviside functions by using the Chan-Vese segmentation model. In order to obtain a better approximation of this histogram and to speed up the calculations, an improved version of the multiphase level set method is introduced. The valleys are then highlighted and isolated by deriving the approximated histogram so that the thresholds are easily extracted by searching the minima of these valleys. Experimental results and a comparative study with three other efficient and known multilevel thresholding methods over synthetic and real images have shown that the proposed method offers very good segmentation results with a low computing time, whatever the complexity of the image and the number of regions composing it.