Study Of Shape Matching Based On Modified Shape Context Algorithm

With the continuous development of computer and digital image technology,the research of image retrieval and matching technology has become a hot topic in the field of pattern recognition,researchers have put forward many excellent methods of image retrieval and recognition.Among them,the shape context algorithm proposed in recent years has been widely used in face recognition and other related fields.But the shape context algorithm mainly uses the contour information to complete the matching operation.Therefore,the shape context algorithm is more sensitive to the shape contour,and it is easy to receive the interference of noise.And there are no clear or more reliable tangents at some sample points,unable to complete shape matching.In daily life,sometimes asking us to quickly get the results of image retrieval,but the matching process of the shape context algorithm is time-consuming,these shortcomings greatly limit the application of shape context algorithm.In this paper,the shortcomings of the shape context algorithm are improved,make it has better robustness.The main contents of this paper are as follows:First of all,according to the shortcomings of shape context algorithm are more sensitive to noise,use the K-nearest filter to preprocess the image contour before extracting the shape contour,this used to reduce the impact of image noise on the robustness of the algorithm.Secondly,on the basis of the existing shape context algorithm,proposes an improved shape context algorithm.In this method,the line between reference coordinates and the centroid of the shape are used as the X axis to establish the log polar coordinates,the problem that some points don't have clear tangent is avoid.The comparison experiments show that the modified shape context algorithm has better robustness than the traditional shape context algorithm.Finally,according to the time-consuming problem of shape context and can't be applied to fast image retrieval,presents a simplified shape context descriptor,and as the input feature of neural network algorithm,to achieve fast image retrieval.This method makes full use of the superiority of neural network algorithm in classification.The time required for image matching is greatly reduced.The experimental results show that the method is suitable for fast image retrieval.The modified shape context algorithm improves the robustness of the algorithm.It can be used for image retrieval and matching,and has good practical value.