An unsupervised evaluation method based on probability density function

Image segmenting is one of the most important steps in movie and image processing and the machine vision applications. The evaluating methods of image segmenting that recently introduced. These evaluation metrics extract some features for each region in a segmented image. In this paper using probabilistic model that utilize the information of pixels (mean and variance) in each region to balance the under-segmentation and over-segmentation. Using this mechanism dynamically set the correlation of pixels in the each region using a probabilistic model. Some famous benchmarks used to test proposed metric performance. Simulation results show this strategy can improve the performance of the unsupervised evaluation segmentation significantly.