Aiming at the problem that the traditional Bayesian model algorithm needs to improve the accuracy of significant region detection, an improved convex hull bayesian model significance detection algorithm is proposed.Firstly, the image foreground is extracted by the popular manifold ranking algorithm as the prior probability of the Bayesian model.Secondly, the color-enhanced Harris corner detection algorithm was used to detect the feature points of the image in the three color Spaces of RGB, HSV and CIELab respectively, and the convex envelope intersection of the three color Spaces was obtained to obtain a reasonable convex envelope, and the probability of observation likelihood was obtained by combining with the color histogram.Finally, according to the calculated prior probability and the observed likelihood probability combined with the Bayesian model, the significance region diagram is obtained.The test results in two public data sets, MSRA-1000 and ECSSD, show that this algorithm has good visual detection effect, and the objective evaluation index F-measure value and accuracy-recall rate curve are superior to the traditional classical algorithm.The validity and accuracy of the proposed algorithm are verified.