Research About Image Retrieval Algorithm Based On The Statistical Model

In the era of digital information,multimedia,which is an important carrier of information transmission and the driving force of modern development,has been widely applied in production and life.In general,it includes two major research areas: database storage system and computer vision.With the continuous improvement of computer storage and computing power,to retrieve valuable information effectively,digital image retrieval,a key technology,has been one of the hot research topic in recent years.From the traditional text-based image retrieval to content-based image retrieval(CBIR),which broke the limitations of image content expression and provided a new direction for investigation.Content-based image retrieval technologies employ the principle of fuzzy matching.In content-based image retrieval system,compared with color,shape,spatial characteristics,texture content information can clearly reflect some image characteristics,such as image smoothing,image density and arrange regulations of pixel array.This paper focuses on image retrieval based on texture feature.The main contents are shown as follows:Firstly,this literature mainly presents the relevant research background,study status and existing problems of image retrieval technology,reviews the basic mechanism,flow chart,framework of content-based image retrieval algorithms,and some image processing technologies.Additionally,this study also presents recent research progress of texture feature extracting technology and its application in relevant fields.Secondly,the basic theory and methods of texture features extracting are also discussed.For the uneven coefficient distribution caused by Dual-Tree Complex Wavelet Transform(DT-CWT),this paper proposes a dual-generalized gauss model,which is a kind of effective feature descriptor and Fully describe the texture information of the image,in which a numerical method is utilized to identification the unknown parameters of model speedily.This method can effectively improve the accuracy of retrieval.Thirdly,within different scale circumstance,there are uncertain relationships between different subband coefficients.This article proposes a method with the combination of the fuzzy set theory and evidence reasoning information fusion(FS-DS).Calculating the uncertain relation on the fuzzy support probability and evaluating the reliability of each scale space on subordinating function.The quantitative operation is carried out in each feature source.The information fusion of each resource can guarantee that the retrieval results are more reliable and accurate that before.Finally,simulation experiments are carried on some widely employed texture images libraries for assessing the performance of algorithms in this study with respective to other classical detection methodologies.Results clearly demonstrate that the double generalized gauss model is able to effectively improve the accuracy of retrieval and optimizer the calculation complexity.In remote sensing image retrieval,multi-scale information fusion method can extract image texture information adequately,and the retrieval precision is also improved significantly.