The Campus Network Bad Image Retrieval Algorithm Design And Implementation

With the development of multimedia and network technology, the application of the image is extensive and content-based image retrieval technique has already become the studied focus. It has combined the technologies, such as image processing, pattern recognition, vision understanding, databases, etc. It is an extremely promising new technology in research and application.In order to analysis the information included in an image, the CBIR system always utilizes the color, texture, shape and other low level image features, to establish the feature vectors as retrieval index. In present time, the main CBIR method is similarity search based on multi-dimension feature vector of image.The main focus of CBIR is not to understand or recognize image. CBIR mainly deals with the ability to retrieve necessary information from the image database in a reasonable time. This kind of query not only utilizes the features of image but also takes into account various technologies, including image processing, image understanding, database, pattern recognition, and man-machine interface.The ideal CBIR aims to retrieve semantic content automatically from images by using image processing and computer vision technologies. Because of the inherent difficulty of images’ semantic, to retrieve semantic content automatically is still an unpractical task. But on one hand there are some inherent connection between images’ visual characteristics and semantic content, e.g., people get the semantic information from the images’visual characteristics, on the other hand, in some specific fields, e.g., medical image diagnosis or oil exploration, the lower visual characteristics such as texture、shape are the condition and criterion for retrieval. Current CBIR often bases on the comparison of the lower visual characteristics.In this thesis, we introduced the background, the state-of-the-art and problems of CBIR at first. Then we gave a deep analysis of the frame of CBIR and related topics engaged in CBIR. such as the extracts of image characteristics, characteristics matching, characteristics description, and image query.Color feature is a common tool of image description, we discuss the model definition, characteristic and suitable range of several color models from optics, visual psychology and digital image processing. Then HSI color model conformed to color cluster feature of human’s vision is established. In the space, a suitable cluster algorithm is used to extract main colors, then the original image is converted to main image.Color characteristics are often represented in the form of histogram. But the color information of image is often very abundant, so it would be computationally intensive to describe Color characteristics in the form of histogram. In this thesis, dominant color is used to describe the color characteristics of image and to perform the task of CBIR.We present an improved method of color space quantification in the extraction of color characteristics so that the quantitative result will better match the visual feature of human.Traditional method of color based image retrieval only performs the matching retrieval according to the global color characteristics of image; the spatial information of images’ color is ignored. In this thesis, we utilized the analysis methods of images texture, presented the concept of dominant color’s texture, extracted images’ feature with dominant color’s texture. The novel algorithm combines the method of analysis texture in color characteristics analysis, thus enjoys the advantages of both methods. Results show that the novel algorithm gets better performance.The analysis of feature of image’s texture is an important technology in CBIR. In this thesis, we analyzed the feature of image’s texture, introduced the methods of texture feature and retrieval of texture feature. Further more, we presented a relatively method to extract feature in frequency space. This new algorithm is clear and computationally simple. We retrieval images in fingerprint image database with this method, results show that the algorithm is very effective and enjoys a high accuracy.Lastly, we designed a CBIR system framework, introduced the workflow and mechanism of this system, gave a detailed analysis of the CBIR database system.