Image Feature Fusion Representation And The Design And Realization Of Matching Algorithm

The image matching algorithm is the core of image matching aided navigation systems. Many algorithms have been proposed to improve the accuracy and real-time performance of image matching. Due to its capability of high-speed information processing and uncertainty information processing, image matching based on feature has attracted considerable attention for image matching in recent years.Comparing with and summarizing the conventional feature-based image matching in detail, adopting the D-S (Dempster-Shafer) theory to image feature fusion representation reasonably, an image feature fusion representation approach from gray value feature, local entropy feature and edge point feature is proposed. The fusion feature has higher stability than any homo-feature. It is show by experiments that the fusion feature improves the precision and probability.In the term of the requirement of image matching performance, a integrated matching algorithm based on wavelet multi-scale properties is proposed and designed. Using the wavelet multi-scale characteristic, the wavelet coefficient maxima are used as the multi-scale edge feature. The matching strategy is the composed of a full searching on the phase of rude matching and thinning layer by layer on the phase of fine matching. The comparability measurement is the maxim correlation algorithm on the rude matching combined with vector angle cosine,which de-noising performance is efficient. At the same time, calculates angle cosine of the corresponding coefficient maxima location vector used as the weight of the maxima vector angle cosine. Implementation of the algorithm is simple. Results are visually pleasant.