Research Of Image Retrieval Algorithm Based On Fuzzy Matrix Learning

With the Development of Multimedia and Network Technology, There Are A Large Number of Images in Our Life. Although the Internet Provides Us with An Inexhaustible Multimedia Information Database, the Problem Is How to Find the Images of User’s Satisfaction Fast and Accurately, It Is A Difficulty Problem of Image Retrieval Currently. So There Is An Urgent Need for Effective Image Retrieval in Many Fields.Content-based Image Retrieval (Cbir) Has Been Proposed and Widely Used the Past Several Years. But the "Semantic Gap" Between the Low-level Features of An Image and High-level Semantic Feature Has Became the Major Obstacle to Cbir Technology. Relevance Feedback Mechanism Narrow the "Semantic Gap" to Some Extent.in This Paper, We First Comprehensively Discuss the Key Technology of Cbir and Relevance Feedback, Then, Made A Deeply Research to Solve the "Semantic Gap", and Provides A High-efficient and Practical Semantic-supported Image Retrieval System. the Main Work of This Paper Include:1. Combine of Color, Texture, Shape and Color-location Feature, It Provide Multi-choice to the User in Order to Meet the User’s Requirement.2. A Relevance Feedback Algorithm Based on Fuzzy Semantic Relevance Matrix Is Proposed. It Update the Value of Fsrm According to User’s Feedback, Which Capture the User’s Retrieval Intention. Through Learning the Value of Fsrm Constantly, and Then Update It to Bridge the "Semantic Gap", Finally, to Improve the Accurate of Retrieval. Experimental Results Prove the Effectiveness of the Algorithm.3. in Order to Improve the Time Performance of Retrieval and Feedback. the Corel Image Database Is Classified Using the Svm. the Method for Svm-based Image Classification Using Color Features. the Experimental Results Show That the Superiority of the Feedback Time on the Basis of the Classification.4. the Learning and Long Learning Algorithm Based on Fsrm Is Proposed. the Learning algorithm can effectively extract the latent semantic information, and update the value of FSRM according to the formula, and the semantic information of FSRM can spread faster to improve the accuracy of feedback. The experimental results prove the efficient of the algorithm. The long learning algorithm is based on the limited feedback learning, then analyze the semantic information of FSRM, and the wrong image in the image class is deleted to further improve the accuracy of feedback.5. Design and implement the retrieval system in this paper with Matlab7.8, it has the function of human-machine interactive. A series of experiments prove the effectiveness of this paper’s algorithm.