Image segmentation based on Differential Evolution algorithm

Threshold segmentation is a critical technology of image segmentation. When the image is low signal-to-noise, the maximum between-cluster variance method (OTSU) cannot provide the ideal result. The 2D maximum between-cluster variance method can perform well with sharply increased computation. This work proposes a new image segmentation method based on OTSU and differential evolution. This solution performs a pre-processing step before the image segmentation. It is shown that differential evolution presents good segmentation result in noisy images. Moreover, the use of this method is easier and faster compared to the 2D maximum between-cluster variance method.