Content-based Image Retrieval And Joint Relevance Feedback Research

In today’s world，with the development of the network technology，content-basedimage retrieval has been a hot research field. but because of the limitation of currentimage understanding technology and the great gap between high level concept and low1evel features for an image，the performance of the CBIR is not good. In order toovercome these difficult problems，human-computer interactive relevance feedbacktechnique is imported into the retrieval and the performance has been improved. In thispaper, the theory and the character of the relevance feedback will be introduced at first,And the SVMbased relevance feedback based on feature weight adjustment andrelevance feedback are discussed, finally the relevance feedback technology applicationprospect is also described.Content-based image retrieval，which directly extracts the visual features of imagesand is able to overcome the subjective text retrieval, fuzzy and the need to manuallymarked shortcomings, is a research hotspot in recent years in the field of image retrieval.To promote the retrieval precision furthermore, this paper integrates the classicalre-weighing relevance feedback algorithm into the UFM algorithm and proposes a newfuzzy region-based relevance feedback algorithm. This algorithm uses the relevantimages and maximizes the weighted product of fuzzy region similarity between thequery image and relevant images to get a weighting vector which assigns weights to thefeatures representing each region and symmetric matrixes to transform the features intonew optimum spaces which the user wants. This paper also combines the SVM basedrelevance feedback algorithm using the global wavelet energy features of both therelevant images and irrelevant images with the proposed fuzzy region-based relevancefeedback algorithm. Experiment shows that the hybrid algorithm performs better.Thisproject focuses on the fuzzy content feature image retrieval and joint relevance feedback,on the base of fuzzy feature, get fuzzy retrieval and relevant feedback technologytogether and joint a variety of commonly used relevant feedback methods to adopt theirstrong points while overcoming some weak points to get a better retrieval result.