Chapter 1: Introduction
In today’s scenario, images are used widely due to the advantages of visual representations. With the continuous development of the networks as well as computers, it is possible to transmit and store large number of images. In recent decades, there is a requirement of image retrieval instead of text retrieval. Content based image retrieval is considered as one of the most efficient and effective ways for accessing visual data (Jabeen et al., 2018). It mainly deals with the content of image like shape, image structure, color rather than annotated text. Over 15 years, CBIR is one of latest and hottest areas of research in context to information retrieval, computing, and multimedia. Number of researchers have implemented algorithms like indexing and retrieval and image mapping whose results indicate that the queries of the visual information are considered as most effective in order to precisely meet with the demands and requirements of users.
Number of CBIR tools and systems have been developed in order to make visual content-based queries. During 90’s, a number of commercial systems were developed in which the QBIC Query by Image Content system is developed by IBM which was helpful in making the queries of the large image databases on the basis of properties of visual image content. These properties include user constructed drawings and sketches, example images, and selected texture and color patterns. Number of online content-based web search engines are available. (Jadhav & Patil, 2012) defines that the image and Advanced Television Lab of Columbia university developed the search engine named as WebSeek. It allows to make queries with the use of examples as well as color composition. Similarly, “Chabot”, which is developed by the Department of Computer Science in University of Columbia allows to search and make queries by searching colors. However, it offers the limited options such as selection of one color at a time.
Apart from the huge development in the CBIR systems, there are number of challenges present in the era of content-based image retrieval. There is still a need for improving the areas like speed of retrieval when working with large number of databases. There is also a challenge in effectively and accurately obtaining the results. However, the simplicity in the retrieval of information is also one of the main requirements. Therefore, following can be the main research questions associated with the content-based image retrieval:
- What are the different techniques for content retrieval?
- What are the features of content-based image retrieval?
- How the different features of CBIR can be obtained from image and how these features can be matched?
- How the retrieval results of CBIR can be represented to the user?
- What is the relation between the accuracy and simplicity of the algorithms?
The research of Content Based Image Analysis, CBIR include two main areas such as database systems and computer vision. Computer vision mainly deals with the image processing, image mapping, and obtaining the descriptors of images. Whereas the database systems mainly focus on database indexing, retrieval and searching. For addressing the formulated research questions, the research mainly focuses on the part of computer vision.
In order to conduct the research, the main focus is to find the techniques which are used in CBIR systems and based on the various features of descriptors. The research aims to provide detailed knowledge about these techniques and methods. The research includes the implementation of effective algorithm in the field of CBIR, which is known as SIFT, Scale Invariant Feature Transform algorithm in order to check its accuracy as well as efficiency.
Chapter 2: Literature Review
(Yue, Li, Liu & Fu, 2011) represents a method in order to extract the texture as well as color features of the image in quick manner with the use of content-based image retrieval. Initially, HSV color space is used as well as quantified. Text features and the color histogram features are used on the basis of the co-occurrence methods and matrix in order to extract the features related to the extracted vectors. The results of the experimental analysis indicate that the fused features retrieval mainly brings the vectors of the images in the single retrieval.
(Jadhav & Patil, 2012) defines an effective content-based image retrieval system on the basis of the evolutionary programming algorithm. The shape, color as well as the texture of the image is extracted for the purpose of query of image and also include the images in the databases. The similar images are retrieved from the databases with the of evolutionary programming algorithm. With the use of precision recall value, the efficiency of the systems is checked.
According to (Jabeen et al., 2018), from the last few decades, the content-based image retrieval has been the hottest research area which represents the viable solution in order to retrieve the images from the image databases. A novel CBIR technique is proposed on the basis visual word fusion that speeds up the robust features. The qualitative as well as quantitative analysis is also performed in order to improve the performance of content-based image retrieval as compared to the fusion features of image descriptors. The results indicate the 78% efficiency of the content-based image retrieval.
Chapter 3: Research Method/Approach
In most of the algorithms of Content Based Image Retrieval, mathematical methods play an important role. However, to access the mathematical solution is difficult and it is impossible to implement practically. Hence, the experimental approach is selected in the research n order to address the research questions. The experimental approach is a method of primary research which is helpful in removing the difficulties of practical implementation. It mainly focuses on the detailed algorithm and clearly identifies the benefits of algorithm. Therefore, with the use of experimental research approach, the algorithm for Shape Based Image Retrieval is implemented for different images and the results are obtained.
Tools and technologies
The research focuses on the experimental implementation of algorithm that include image processing. The algorithm is implemented with the use of Microsoft .net framework platform as well as with the use of C# and GDI+ programming languages. In order to provide managed interface for the GDI+, .net framework is used. hence, it can be easy to process the images. The Microsoft visual studio .net is used as the IDE. The results of the experimental analysis provide the usefulness of .net framework library as well as C# language for the image processing.
- Jabeen, S., Mehmood, Z., Mahmood, T., Saba, T., Rehman, A., & Mahmood, M. (2018). An effective content-based image retrieval technique for image visuals representation based on the bag-of-visual-words model. PLOS ONE, 13(4), e0194526. doi: 10.1371/journal.pone.0194526
- Jadhav, S., & Patil, V. (2012). An effective content Based Image Retrieval (CBIR) system based on evolutionary programming (EP). 2012 IEEE International Conference On Advanced Communication Control And Computing Technologies (ICACCCT). doi: 10.1109/icaccct.2012.6320793
- Yue, J., Li, Z., Liu, L., & Fu, Z. (2011). Content-based image retrieval using color and texture fused features. Mathematical And Computer Modelling, 54(3-4), 1121-1127. doi: 10.1016/j.mcm.2010.11.044