Motion segmentation method of two-dimensional view image scene

The invention relates to a motion segmentation method of a two-dimensional view image scene in the technical field of image processing. The method comprises the following steps of: carrying out local feature extraction and feature abstract description; carrying out preliminary matching on extracted local features to form a feature point pair matching set; generating an initial motion model for each feature point pair to obtain an initial motion model set; mapping each initial motion model into a high-dimensional probability vector; appointing a weight for each high-dimensional probability vector; obtaining a plurality of main motions through guide sampling and partition processing and ensuring that residual probability vectors in a probability vector set are smaller than a probability threshold value; and carrying out the subsidiary appointment of the feature point pair matching and rejecting abnormalities. The method does not need any priori knowledge about motion model quantities and can be used for processing a large quantity of motion models under the condition of no need of a large quantity of feature points and better processing noise data, solves the problem of limitation of mean shift, enlarges the application range and obtains certain improvement in the aspect of time consuming.