A Comparative Study Of Feature Extraction Techniques In Content-based Image Retrieval

In the information era,there has been a huge amount of image information in the life.It is the purpose of image retrieval technology to retrieve the similar image from the massive information.The previous image retrieval technologies are mainly based on text.Subsequently,an algorithm based on the color,texture,shape and other features of the image are proposed,which is based on the description of semantic content,it is called content-based image retrieval.The difficulty and hotpots in content-based image retrieval technology is still feature extraction technology.The first part of the paper discusses the background and current situation of this topic,and the content-based image retrieval technologies are introduced.And then the effect of two traditional feature extraction techniques in image retrieval is analyzed experimentally.Finally,three kinds of better feature extraction methods used in image retrieval are introduced and compared in detail.The main works are as follows:(1)In the traditional feature extraction methods,this paper mainly studies the feature extraction method based on HSV space color and the feature extraction method based on gray level co-occurrence matrix in CBIR.And we analyzed the result of the two traditional retrieval algorithms in image retrieval.(2)The application of hash algorithm in image retrieval system is studied.In the hash algorithm,the main research is the mean hash algorithm feature extraction technology,and through the discrete cosine transform instead of image size reduction to improve it.Then,the improved mean hash algorithm and the improved mean hash algorithm are applied to image retrieval,and the retrieval results are compared.Experiments show that the improved mean hashing feature extraction technology is better than the unimproved technology and two traditional feature extraction techniques in CBIR.(3)The application of SIFT algorithm in image retrieval system is studied.This paper introduces the basic principle of SIFT algorithm to extract the feature descriptor and the concrete step,and through the experimental analysis of the SIFT feature extraction method in CBIR.(4)The application of convolution neural network in image retrieval system is studied.Through the study of classical convolutional neural network model,to extract image features and proposed a new type of pre-training convolutional neural network,and through the experimental analysis of the retrieval effect of the convolutional neural network models based on CBIR.In order to compare the performance of various feature extraction methods,the same experimental conditions are used in various algorithms.Experimental results show that compared with the traditional feature extraction techniques,the new feature extraction technique has better retrieval performance.