A Modification of Feature Extraction and Matching Algorithm Based on the Points of Interest

This thesis mainly study the problem of the accuracy and optimization of the feature extraction on the points of interest in images. Due to the complexity of the image and the interference of noise in the image, the traditional feature extraction algorithm based on the points of interest is difficult to extract the information which users interested in. In order to solve the above problems, an improved feature extraction algorithm based on the points of interest is proposed. Firstly, the points of interest of groups can be used to form geometric invariant description regions of the image. The re- sampling images can by made by all the feature points, which is set by the points of interest in the image. And then take the transform operation to the image by the Hough transform method. This approach means that each match is equivalent to the local two-dimensional change. The simulation results show that the improved algorithm can effectively extract the useful information of the image effectively, and saves execution time, reduces the complexity.