Image Retrieval Using Ann

Only available on StudyMode
  • Topic: Fuzzy logic, Artificial neural network, Content-based image retrieval
  • Pages : 16 (3358 words )
  • Download(s) : 121
  • Published : December 1, 2012
Open Document
Text Preview
Image Retrieval Based on Color and Texture Feature Using Artificial Neural Network Syed Sajjad Hussain#1, Manzoor Hashmani#2, Muhammad Moin uddin#3 #

Faculty of Engineering, Sciences and Technology, IQRA University, Karachi 1

engr.sajjadrizvi@yahoo.com, 2mhashmani@yahoo.com, 3mmoin73@yahoo.com Abstract. Content-based image retrieval CBIR is a technique that helps in searching a user desired information from a huge set of image files and interpret user intentions for the desired information. The retrieval of information is based on features of image like colour, shape, texture, annotation etc. Many of the existing methods focus on the feature extraction and to bridge up the gap between low level features and high level semantics. In this paper we propose a supervised machine learning (SML) using artificial neural network (ANN) and singular value decomposition (SVD) for image retrieval. Specifically we use back propagation algorithm (multilayer perceptron) (MLP) for training and testing our proposed model. Experimental results show that by changing parameters of feature vector back propagation method can have 62% precision instead of 49% as claimed by in Hyoung Ku LEE, Suk In Yoo [1]. Keywords Colour based image retrieval, back propagation algorithm, artificial neural network based image retrieval, multilayer perceptron (MLP).

1

Introduction

Previously the information was primarily text based. But with the rapid growth in the field of computer network and low cost permanent storage media, the shapes of information become more interactive. The people are accessing more multimedia files than the past. In past, images, videos and audio files were only used for the entertainment purpose but nowadays these are the major source of information. Because of intense dependency on multimedia files for information searching, to obtain a desired result is a major problem as the search engine searches within the text associated with the multimedia files, instead of the contents of information. Intelligent and optimized text searching has already matured but there is a gigantic space available for the intelligent and optimized Contents Based Image Retrieval (CBIR). In this age of information technology we are intensively depending on the visual information and its contents as well as with the rapid growth of human computer interaction (HCI). This information requires new methods to archive and access multimedia information. While conventional databases allow for textual searches on metadata only[2]. In applications like medical image management, multimedia libraries, document archives, art collections, geographical information systems, law enforcement agencies, education, websites are surrounded by rich and un-optimized

2

text description that may provide just contextual hints or semantic information about the contents. But the optimization cannot be achieved with this rich text description because most of the text is not relevant with the contents information. Moreover, the traditional keyword-based approaches require an enormous amount of human effort for manual annotation. In multimedia applications, comparisons often are not based on exact match, but rather based on similarity comparison. Therefore, there is an essential need of optimized and intelligent retrieval. The commercial image search engines available as on date are: QBIC, VisualSeek, Virage, Netra, PicSOM, FIRE, AltaVista, etc. Region-Based Image Retrieval (RBIR) is a promising extension of CBIR [3]. Almost all the CBIR systems designed so far widely use features like color, shape and texture [4]. In this paper, we propose a Neural Network-based Image Retrieval System (NNIRS) specifically Back Propagation (BP) for training and testing. Initially all the training images are converted into feature vector (FV) containing color and texture feature. FV vector of each image is recorded in feature vector table (FVT) that is used for training and adjusting the...
tracking img