Texture Image Analysis Based On Trace Transform

With the rapid development of artificial intelligence,image has become the main media of communication in the Internet world.The texture is one of the important visual features of the image,it contains rich information,and has high research value,too.Texture analysis is a subject of texture image processing and texture classification is one of the important direction of the research.The most important index to influence the classification performance is the discriminative ability of extracted feature information.However,there is no precise and uniform definition of texture,which leads to more challenging problems in texture analysis,and a variety of feature extraction algorithms are also derived.Trace transform is a novel and effective image feature extraction algorithm,which can extract rotation,scaling and translation invariant features.It has become one of the research hotspots in the field of image processing and computer vision.Trace transform can be used to obtain image features by triple-functional cascading operation.In the process,some information of the image is inevitably lost,more groups of functional will be needed,and there is a problem of insufficient real-time performance.In conclusion,it has certain theoretical and practical value to improve it.On this basis,the research contents and achievements in this paper are as follows:(1)In this paper,we analyze the main feature extraction algorithms,analyze the principle of Trace transform,and make a deep research on its performance.(2)In trace transform,there is a problem that the image features extracted by it lack description of texture edge information,and the computation cost is high,too.Therefore,a new fusion feature extraction algorithm,multi-resolution trace transform,is proposed.Firstly,the wavelet transform was introduced in trace transform,low frequency feature sub images and high frequency edge sub images of texture images at different frequencies are obtained by using nonsubsampled wavelet transform,which overcomes the deficiency of insufficient image texture feature description in trace transform.Then,we carried out a set of functional trace transform on each level sub images to obtain the fusion features of texture image,which not only obtains the edge information of the image,but also avoids the problem of high cost.Finally the fusion features were sentinto support vector machines to classify the images.The experimental results show that the fusion features of multi-resolution trace transformation have better ability to describe image texture than the original Trace transform,and achieves higher average classification accurate rate in the texture image library,the time efficiency is greatly improved compared to the traditional Trace transform,it has certain advantages compared with other algorithms in classification performance too.(3)Trace transform as a global feature extraction algorithm,combined with local features is also one of the direction of its in-depth study.As an effective local texture feature description operator,local binary pattern can be combined with global features to obtain more informative features,and it has been successfully applied in the field of image processing.In view of this,a multi-feature fusion algorithm based on Trace transform and LBP is proposed and successfully applied to the fracture image classification.Firstly,the global texture features of the image are extracted by the Trace transform,the local texture features are extracted by the local binary pattern,and then dynamic weighted discriminant energy analysis is used for feature selection and adaptive weighted fusion.Finally,the classification is carried out by SVM.Experiment results on the image database of metal fracture show that the proposed method has high recognition rate,which has obvious advantages compared with other algorithms.The proposed algorithm also has good recognition rate in other texture database,so it has good generalization ability.