Abstract Image segmentation is typically used to detect object contours in an image. In this article, a threshold-gradient based image segmentation method (TGISM) is proposed to extract objects from an image, based on the techniques of object contour gradient, gradient vector
flow, and thresholding. In this article, a thresholding method minimizing within-class variance is also presented. The experimental results validate that the TGISM can give better segmentation results than the results obtained by the existing set of methods reported in the literature. In this
article, a genetic algorithm based on accumulated historical data is proposed for determining the appropriate values of the parameters used in the TGISM.