Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/9091
Title: | Content based image retrieval |
Researcher: | Thepade, Sudeep |
Guide(s): | Kekre, H B |
Keywords: | Content base Image Retrieval |
Upload Date: | 23-May-2013 |
University: | Narsee Monjee Institute of Management Studies |
Completed Date: | 2011 |
Abstract: | Modern image search engines retrieve the images based on their visual contents, commonly referred to as Content Based Image newlineRetrieval (CBIR) systems. Typical CBIR systems can organize and newlineretrieve images from image databases, automatically by extracting newlinesome features such as color, texture, shape from images and newlinelooking for similar images which have similar feature. One problem newlineof this approach is reliance on visual similarity to judge semantic newlinesimilarity, which creates problems due to semantic gap between newlinelow-level content and high level concepts. Even with the subsistence newlineof this problem, if aggressive attempts are made CBIR can be used newlinefor real life applications. For example in spite of the open problems newlinelike robust text understanding, Google and Yahoo have become newlinemost popular for searching. newlineThe work presented here mainly focuses on efficient CBIR methods newlinewith help of representation of converting the visual content of newlineimages in feature vector using proposed techniques. The proposed newlineCBIR methods using Colour, Transformed Image, Texture and newlineShape content are proved to be better and faster using test bed of newline1000 variable size images spread across 11 image categories. newlineIn consideration of colour content as feature, the proposed newlineapproaches of using image colour averages and block truncation newlinecoding for CBIR are proposed. Image averaging techniques are newlinebased on taking averages of colour content of image, which can be newlineconsidered as feature vector for image retrieval. The image pixel newlinedata can be represented in form of the feature vectors with reduced newlinedimensions as row mean (RM), column mean (CM), forward newlinediagonal mean (FDM), backward diagonal mean (BDM). Use of the newlinecolour averaging techniques helps in obtaining faster and better newlineimage retrieval techniques. The FDM has been observed to give best newlineperformance among these colour averaging based CBIR methods.Image tiling is dividing image into equal and non-overlapping newlinesquare parts. |
Pagination: | 311p. |
URI: | http://hdl.handle.net/10603/9091 |
Appears in Departments: | Department of Computer Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 115.15 kB | Adobe PDF | View/Open |
02_table of content.pdf | 162.17 kB | Adobe PDF | View/Open | |
03_abstract.pdf | 105.72 kB | Adobe PDF | View/Open | |
04_chapter 1.pdf | 146.05 kB | Adobe PDF | View/Open | |
05_chapter 2.pdf | 162.29 kB | Adobe PDF | View/Open | |
06_chapter 3.pdf | 984.8 kB | Adobe PDF | View/Open | |
07_chapter 4 a.pdf | 706.42 kB | Adobe PDF | View/Open | |
08_chapter 4 b.pdf | 1.24 MB | Adobe PDF | View/Open | |
09_chapter 5.pdf | 842.82 kB | Adobe PDF | View/Open | |
10_chapter 6.pdf | 480.54 kB | Adobe PDF | View/Open | |
11_chapter 7.pdf | 409.8 kB | Adobe PDF | View/Open | |
12_references.pdf | 229.84 kB | Adobe PDF | View/Open | |
13_appendix i - ii.pdf | 6.09 MB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: