Activecontour model based on Bayesian minimum misjudgment

In order to improve the segmentation accuracy of images disturbed by noise and texture, an active contour model based on Bayesian minimum misjudgment is proposed. Firstly, the Mahalanobis distance between the target and the background is used as the decision rule of the misjudgment probability, and the PDE equation of directional diffusion is used to evolve the observed image, so that the diffusion occurs only along the tangent direction, and a smooth approximate image of the original image is obtained, thus the Mahalanobis distance between the target and the background increases and the misjudgment probability decreases. Then, the active contour model fitting is conducted in combination with local binaries, and the feature function is iteratively evolved until it converges to the target edge. Finally, the model is solved numerically by the method of gradient descent and iterative convolution thresholding. The experimental results show that the proposed model can well eliminate the influence of interference information, accurately extract the target contour, and reduc the probability of misjudgment. Compared with several classical variational image segmentation models, the values of DSC and IOU indicate that the proposed model has better experimental results.