Research On Solar Image Retrieval Based On Curvelet Feature Application

In recent years,as the image libraries are constantly increasing,digging the contents of large image libraries(such as solar image)is a challenging work.Therefore,it is becoming increasingly difficult to find specific images in the relevant image library.People want to get specific images,but these images have been buried in the large image libraries.For example,it is still not a simple task to get a specific type of photos from solar image library which has tens of thousands of images.The search engine used in the retrieval system must be both efficient and efficient.And the content-based image retrieval technology(CBIR)fulfills these requirements.CBIR make use of the color,texture,shape,edge and other underlying visual features of image to describe the image content and carry out the search work.CBIR system overcomes many shortcomings of the text-based image retrieval methods,which attract the researchers' attention.It is undoubtedly a good choice to describe the solar images using of texture features.In the past,wavelet transform is widely used because of the ability of time and frequency domain analysis.But wavelet transform does not always have an advantage at every field.So when dealing with hyperplane singularity,wavelet transform is not so suitable.As a new instrument of analyzing,curvelet can represents an image with straight line feature sparsely,and it can also be applied to two-dimensional image processing process.On the basis of the study of ridgelet transform and curvelet theory,this paper carried out research and analysis of curvelet transform implementation process.The paper mainly does the following work:First,starting from CBIR system,when does the retrieval this paper selected the texture feature as the retrieval feature,selected Euclidean distance as the similarity measure method,and selected precision and recall curve as evaluation standard.Second,this paper applied the curvelet texture feature to the solar image retrieval process,and made use of the retrieval experiment to virify the performance.And this paper also carried out the optimization of curvelet feature extraction algorithm,improved the retrieval efficiency.Third,this paper also carried out the weighted fusion of the curvelet feature and wavelet feature,the feature of the weighted fusion also act well in the image retrieval.