Research On Image Clustering And Retrieval Based On Image Underlying Features

At present,the development of new technologies in computer and Internet is changing rapidly.These technologies have greatly facilitated our life and brought many new problems.Digital cameras,produce a large number of digital images,and cell phone photography makes the number of images increase.How to store and manage these digital images and how to accurately and quickly find the images that meet our requirements when needed,these problems gradually attract people's attention.There have been some ways to solve these problems since the 90 s of twentieth Century,Content Based Image Retrieval is one of the most widely concerned methods in these methods,with the maturity and development of this technology,this technology has become a hot spot to solve the problem of image retrieval.In the early stage of image retrieval,there is no good use of the features of the image content,there is subjectivity of the image between the visual angle and the image means,this is what people often call the so-called semantic gap,it is a problem that needs to be solved in the process of image retrieval research and Application.Using relevance feedback technology,we can analyze users' requirements in the interaction between the user and the system,and gradually eliminated the results of the non-qualified retrieval,and ultimately leaving the eligible picture to get a relatively satisfactory result,which is an effective way to narrow the semantic gap.This article is mainly done in the following aspects:1?The low layer feature of the image is extracted and the similarity is matched.Low level feature extraction is the basis for the realization of content-based image retrieval technique,this paper we will discuss the image color feature,shape feature,texture feature and the extraction method.We will extract the underlying feature from the image in the general standard image library and quantify the extracted data to form a database.After extracting the basic features of the image,at last,we will discuss the similarity of image features.2?The application principle of K-means in image clustering is discussed.Based on the traditional K-means algorithm,the positive feedback strategy is introduced to adjust the clustering center to cluster,overcome the problems in traditional Kmeans algorithm,and improve the accuracy of image clustering.3?The use of positive feedback strategies in image retrieval,based on the results ofcertain features of the image are weighted to highlight these features makes the user feedback,with the same characteristics of the image from the image of irrelevant collection emerges,ultimately meet user requirements of the image.