P.Sankara Rao$1 E.Vamsidhar#2. G.Samuel Vara Prasad Raju$3 Ravikanth Satapati$1 KVSRP.Varma$1. $1 Asst Professor, Dept. of CSE, GITAM University, ...
P.Sankara Rao.et. al. / International Journal of Engineering Science and Technology Vol. 2(4), 2010, 559-563
AN APPROACH FOR CBIR SYSTEM THROUGH MULTI LAYER NEURAL NETWORK P.Sankara Rao$1 E.Vamsidhar#2 G.Samuel Vara Prasad Raju$3 Ravikanth Satapati$1 KVSRP.Varma$1 $1
Asst Professor, Dept. of CSE, GITAM University, AP, India. #2 Lecturer, Dept. of IT, GITAM University, AP, India. $3 Associate Professor in Computer Science, Andhra University, AP, India. Abstract: In these days an image retrieval system has become a challenging task. Many systems based on the text based retrieval but the need of image based retrieval system that takes an image as the input query and retrieves images based on image content is more complicated task. Content Based Image Retrieval is an approach for retrieving semantically-relevant images from an image database based on automatically-derived image features. The exclusive aspect of the system is the utilization of hierarchical and multilayer network which includes RBFN . The proposed procedure consists of two stages. First, here we are going to filter most of the images in the hierarchical clustering and then apply the clustered images to RBFN network, so that we can get better favored image results. Key words: CBIR, Clustering, RBFN, Hierarchical clustering, RGB, Similarity. Introduction This section gives an introduction to content based image retrieval system (CBIRS) and the technologies used in them. Image retrieval has been an extremely active research area over the last 10 years, but first review articles on access methods in image databases appeared already in the early 80s[1]. Enser [2] gives an extensive description of image archives, various indexing methods and common searching tasks, using mostly text based searches on annotated images. In [3], an overview of the research domain in 1997 is given and in [4], the past, present and future of image retrieval is highlighted There are several reasons why there is a need for additional, alternative image retrieval methods apart from the steadily growing rate of image production. It is important to explain these needs and to discuss possible technical and methodological improvements. Image retrieval is the process of browsing, searching and retrieving images from a large database of digital images. Advances in data storage and image acquisition technologies have enabled the creation of large image datasets. In order to deal with these data, it is necessary to develop appropriate information systems to efficiently manage these collections. It is simple to identify a desired image from a small collection simply by browsing, but we need more effective techniques with collections containing thousands of items. Image searching is one of the most important services that need to be supported by such systems. In general, two different approaches have been applied to allow searching on image collections: one based on image textual metadata and another based on image content information. The first retrieval approach is based on attaching textual metadata to each image and uses traditional database query techniques to retrieve them by keywords [5,6]. However, these systems require a previous annotation of the database images, which is a very laborious and time-consuming task. Furthermore, the annotation process is usually inefficient because users, generally, do not make the annotation in a systematic way. In fact, different users tend to use different words to describe a same image characteristic. The lack of systematization in the annotation process decreases the performance of the keyword-based image search. Image retrieval systems have not kept pace with the collections they are searching. The shortcomings of these systems are due both to the image representations they use and to their methods of accessing those representations to find images. The problems of image retrieval are becoming widely recognized, and the search for solutions an increasingly active area for research and development. In recent years, with large scale storing of images the need to have an efficient method of image searching and retrieval has increased. It can simplify many tasks in many application areas such as fingerprint identification biodiversity information systems, digital libraries, crime prevention, medicine, historical research, artificial intelligence, military, education, web image searching. Content-Based Image Retrieval (CBIR) systems [7-9] shown in Fig-1.In these systems, image processing algorithms (usually automatic) are used to extract feature vectors
ISSN: 0975-5462
559
P.Sankara Rao.et. al. / International Journal of Engineering Science and Technology Vol. 2(4), 2010, 559-563 that represent image properties such as color, texture, and shape. In this approach, it is possible to retrieve images similar to one chosen by the user (query-by-example). There by we can overcome the disadvantages of the text based retrieval systems .The main advantages of this approach is the possibility of an automatic retrieval process, contrasting to the effort needed to annotate images.
Fig 1.Typical Architecture of Content Based Retrieval System
This system distinguishes the different regions present in an image based on their similarity in color, pattern, texture, shape, etc. and decides the similarity between two images by reckoning the closeness of these different regions. The CBIR approach is much closer to how we humans distinguish images. Thus, we overcome the difficulties present in text-based image retrieval because low-level image features can be automatically extracted from the images by using CBIR and to some extent they describe the image in more detail compared to the textbased approach .Image clustering inherently depends on a similarity measure, image categorization has been performed by varied methods that neither require nor make use of similarity metrics. Image categorization is often followed by a step of similarity measurement, restricted to those images in a large database that belong to the same visual class as predicted for the query. In such cases, the retrieval process is intertwined, whereby categorization and similarity matching steps together form the retrieval process. Similar arguments hold for clustering as well, due to which, in many cases, it is also a fundamental “early” step in image retrieval [10]. We have used Hierarchical and the RBFN network procedures which can be applied for scalable image retrieval from large databases.. Hierarchical Algorithm is used to group similar images into clusters to increase the retrieval speed. RBFN network is a multilayer neural network approach to retrieve the desired image which is run on Kmeans clustering algorithm. CBIR systems based on features like color, shape, texture, spatial layout, object motion, etc., are cited in[11],[12].Color is one of the most widely used features for image similarity retrieval, Color retrieval yields the best results, in that the computer results of color similarity are similar to those derived by a human visual system that is capable of differentiating between infinitely large numbers of colors. One of the main aspects of color feature extraction is the choice of a color space. A color space is a multidimensional space in which the different dimensions represent the different components of color [13]. Most color spaces are three dimensional. Example of a color space is RGB, which assigns to each pixel a three element vector giving the color intensities of the three primary colors, red, green and blue. The space spanned by the R, G, and B values completely describes visible
ISSN: 0975-5462
560
P.Sankara Rao.et. al. / International Journal of Engineering Science and Technology Vol. 2(4), 2010, 559-563 colors, which are represented as vectors in the 3D RGB color space. As a result, the RGB color space provides a useful starting point for representing color features of images. Proposed Method We have used the hierarchical clustering algorithm to group the images into clusters based on the color content. And then the output group was assigned to the RBFN network the primary reason for this is that the Learning process in RBFN network has two stages and both stages can be made efficient by appropriate learning Algorithms. Although finding the optimal k clusters for a set of data is NP-complete, k-means is quite efficient and generally produces good results. Especially their middle layer composed of receptive fields, using k-means clustering technique The use of RBFN is to retrieve the desired image with the comparison with the query and the database, especially the number of input neurons number of hidden neurons. Centers ci, width σi and weight wi and the use of K-means clustering technique to analyze and find Clusters in the training data. The results of this grouping are establishing prototypes of the receptive fields. The proposed RBFN network model is shown in Fig 2. The activation function in the hidden layers is Gaussian function represented by O(x) = exp (-x2 ).
Figure 2: RBFN Network model
HIDDEN LAYER: O(x) = exp (-(||x-c||2 ) / σ2 ) OUTPUT LAYER: Σ O(x) Wi The weights W1, W2…Wn denotes the weights of the arcs from the hidden layers nodes to the output node. Note that the initial values of W1, W2 ….Wn are adjusted so that the nodes in the hidden layer have same contribution to the Output node . Using the hierarchical clustering algorithm we obtained a groups of similar images V1,V2 and by using RBFN neural network with the K-means clustering algorithm, we can get a retrieved image by comparing with the query image. The retrieval image is compared against the query image. Assume for a certain set of inputs the retrieval image of the network as RI. Furthermore QI denotes the Query image given by the user. The error is estimation is determined by Error = QI – RI which is used to determine the error of nodes in the hidden layer. In the output layer: 1: Adjust weight by Wi (new) = Wi (old) +ΔWi 2: Weight change is compute by ΔWi=δ. Error O(x)
ISSN: 0975-5462
561
P.Sankara Rao.et. al. / International Journal of Engineering Science and Technology Vol. 2(4), 2010, 559-563 Where
δ= learning rate Error= QI – RI O(x) = output of Activation function in hidden layer
3: Repeat iteration until Convergence. Here we are going to filter most of the images in the hierarchical clustering and then apply the clustered images to RBFN network, so that we can get better favored image results. Brief details on the implementation of these two clustering algorithms are presented below.
Figure 1: Block Diagram for proposed Image Retrieval System
Algorithm: Step 1: Input the Image and perform Hierarchical clustering. Step 2: Consider the Every point as its own cluster. Step 3: Find Most Similar Pairs of Clusters. Step 4: Merge those two points to one parent cluster. Step 5: Repeat Step 3 to Step 5 until all points are merged into one cluster. Step 6: Apply RBFN network to the required image set obtained from Hierarchical clustering. Step 7: Enter How Many Clusters (Let “k”). Step8: Randomly Guess K Cluster center Locations. Step 9: Each Data point finds out which center it’s closest to. Step 10: Thus Each Center “Owns” Set of Points. Step 11: Each Center Finds the Centroid of its Own Points. Step 12: Center now moves to the New Centroid. Step 13: Repeat Step 9 to Step 12 Until Terminated. Retrieval accuracy with RBFN Network Clustering is a mutually exclusive partitioning process of the feature space of feature vectors in a meaningful way for the application domain context. With the clusters, we may perform nearest neighbor search efficiently. The unique aspect of this system is the utilization of hierarchical and RBFN network techniques. Here we are going to filter most of the images in the hierarchical clustering and then apply the clustered images from the hierarchical clustering to RBFN Network, so that we can get better favored image results. After hirecheracal clustering it is assigned to the RBFN Network the clustered centriods which are obtained by the K-Means clustering algorithm are assigned to the Gaussian function in the hidden layer the output of the hidden layer are multiplied with the weights W1, W2…Wn . The result is the retrieval image is compared with the query image. The retrived image will be more accurate by adjusting the weights between hidden layer and the output layer
ISSN: 0975-5462
562
P.Sankara Rao.et. al. / International Journal of Engineering Science and Technology Vol. 2(4), 2010, 559-563 Conclusion A new approach has been presented in this paper for Content Based Image Retrieval using Hierarchical and RBFN network techniques where images are initially clustered into groups having similar color content and then the preferred group is clustered using Hierarchical clustering which assists for faster image retrieval and also allows the search for most relevant images in large image databases. RBFN Network is a multi layer neural network approach which uses K-Means clustering and Gaussian function to retrieve the similar images by comparing the images in the databases and query images. This work can be extended by integrating with Fuzzy C-means clustering algorithm for better efficiency. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]
S-K Chang,T.Kunii Pictorial database applications, IEEE Computer 14 (11)(1981) 13-21. P. G. B. Enser, Pictorial information retrieval, Journal of Documentation 51 (2) (1995) 126-170. A. Gupta, R. Jain, Visual information retrieval, Communications of the ACM 40 (5) (1997) 70-79. Y. Rui, T.S.Huang, S.-F.Chang, Image retrieval Past, present and future, in: M. Liao(Ed.),Proceedings of the International Symposiumon Multimedia Information Processing Taipei,Taiwan,1997. V.E. Ogle andM. Stonebraker. Chabot: Retrieval from Relational Database of Images. IEEEComputer, 28(9):40–48, Sep 1995. H. Lieberman, E. Rosenzweig, and P. Singh. Aria: An Agent for Annotating and RetrievingImages. IEEE Computer, 34(7):57–62, 2001. M. Flickner, H. Sawhney, W. Niblack, Q. Huang J. Ashley, B. Dom, M. Gorkani, J. Hafner,D. Lee, D. Petkovic, D. Steele, and P. Yanker. Query by Image and Video Content: theQBIC System. IEEE Computer, 28(9):23–32, Sep 1995. A. W. M. Smeulders, M. Worring, S. Santini, A. Gupta, and R. Jain. Content-Based Image Retrieval at the End of the Years. IEEE Transactions on Pattern Analysis and MachineIntelligence, 2(12):1349–1380,December 2000. Y. Rui, T. S. Huang, and S. F. Chang. Image Retrieval: Current Techniques, PromisingDirections, and Open Issues. Journal of Communications and Image Representation,10(1):39–62,March 1999. Ritendra Datta, Dhiraj Joshi, Jia Li, And James Z. Wang, “Image Retrieval: Ideas, Influences, and Trends of the New Age”, ACM Computing Surveys, Vol. 40, No. 2, Article 5, Publication date: April 2008. V.N.Gudivada and V.V.Raghavan.: Special issue on content-based image retrievalsystems - guest eds. IEEE Computer. 28(9) (1995) 18-22 M.De Marsicoi, L.Cinque, and S.Levialdi.: Indexing pictorial documents by theircontent: a survey of current techniques. Image and Vision Computing. 15(2) (1997) 119-141 Daniela Stan & Ishwar K. Sethi, Image Retrieval Using a Hierarchy of Clusters. 237-241.
ISSN: 0975-5462
563
