A Research Of Feature Coding Model And Algorithms In Image Classification

The popularity of multimedia technology makes millions of pictures appear every day and image resources increase more and more.Digital image is one of the most important carriers of information.Compared with other carrier such as text information,digital image is more specific and direct on the expression of the content.For this reason,people pay more attention to digital pictures.However,with the increasing of image database,it's more difficult to efficiently find which picture we need.Therefore,image classification is becoming more and more important.The spatial pyramid model is the common model in image classification,and the model segments each layer of the image according to certain rule,then it calculates its local feature histogram in the sub-region.Because of getting more image information,the spatial pyramid model gets an advanced performance in image classification.Based on the performance of the spatial pyramid model,this paper makes it as the main framework for research.The main contents are as follows:First,this paper analyzes some coding algorithms such as sparse coding algorithm and local constraints linear coding algorithm.Also,this paper introduces some algorithms that are related to image classification on the basis of spatial pyramid model.Next,this paper proposes a new algorithm–Multi Feature Coding Algorithm based on locally constrained linear coding algorithm and non-negative sparse coding algorithm with L₂ norm.The algorithm combines the non-negative sparse coding algorithm with L₂ norm and the local constrained linear coding algorithm.Compared with local constrained linear coding algorithm and non-negative sparse coding algorithm based with L₂ norm,the algorithm effectively combines the advantages of the two coding algorithms.Last,in order to verify the effectiveness of the algorithm,experiments are carried out on three data sets.Experimental results show that the proposed algorithm achieve better classification accuracy compared with nonnegative sparse coding algorithm with L₂ norm and local constraints linear coding algorithm.The experiment of this paper is based on the basic framework of the SPM model.And the LibSVM Package is used in the image classification phase.In order to compare accuracy of the algorithms,a lot of experiments are carried out on the dataset of Caltech-101,Scene-15 and UIUC-Sports.The experimental results show that the proposed model based on multi feature coding algorithm has higher classification accuracy.