grp6 - CONTENT BASED IMAGE RETRIEVAL Presented by Sandeep...

Info iconThis preview shows pages 1–8. Sign up to view the full content.

View Full Document Right Arrow Icon
CONTENT BASED IMAGE RETRIEVAL Presented by Sandeep Kumar Jain-2008JCA2058 Sireesha Madabhushi-2008JCA2059 Sritej Puvvada-2008JCA2066 Yugandhar Naik M-2008JCA2080
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
OUTLINE OF THE PRESENTATION • Introduction to CBIR and steps that led to its evolution. • An algorithm to capture texture feature from an image using curvelet transform. • An algorithm for color extraction from an image using DCTVQIH DCTVQIH. • Explanation of image indexing using K tree.
Background image of page 2
INTRODUCTION • Search engines such as Google retrieve images based on the text associated with them and do not look at the content of the image. • Try getting images of say ‘waterfall’ from Google. Observe that as you move on from page to page, the number of irrelevant images you see get increased number of irrelevant images you see get increased (typically from 20 th page or so). • You can in fact fool the Google search engine to believe that a photo of your dog depicts a waterfall!!!!! Problem – Google does not understand your image. Retrieval procedure is based on metadata (text) of the image rather than its actual content.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
WHY CBIR IS IMPORTANT? If we continue with image retrieval based on the text annotated with images, there are a lot of drawbacks. • Text annotation is subjective and varies across different cross sections of people cross sections of people. • Annotating a huge database manually is a very humungous task. Computer should work for us on our behalf and not the other way round (users should not be required to annotate the image database with text)
Background image of page 4
PARADIGMS OF CBIR There are 3 paradigms for CBIR. • Query By Visual Example (QBVE) – User provides an example image as a query. Images found similar to this are returned. Semantic Retrieval – Here also a query image is Semantic Retrieval – Here also a query image is provided but the resultant images are found by using the semantic content of the query image. • Query By Semantic Example (QBSE) – Makes use of a training set of images judiciously. Can include relevance feedback from user.
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
QUERY BY VISUAL EXAMPLE • Retrieves images by strict visual similarity. • Ranks database images by a similarity to the user provided image. • The system extracts a signature from the query, compares this to that of the images stored in the compares this to that of the images stored in the database. • A basic approach for composing image signatures includes matching of color histograms. • Modern representations rely on more sophisticated algorithms for optimal retrieval performance.
Background image of page 6
MINIMUM PROBABILITY OF ERROR (MPE) RETRIEVAL Figure 1. Minimum probability of error retrieval. (a) MPE retrieval architecture.The system decomposes images into bags of local features and characterizes them by their distributions on the feature space. Database images are ranked by posterior probability of having generated the query features. (b) Retrieval results. Each column shows the three best matches (among 1,500) to the query image shown at the top.
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 8
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 05/22/2011 for the course COMP 207 taught by Professor Zhangli during the Spring '11 term at University of Liverpool.

Page1 / 48

grp6 - CONTENT BASED IMAGE RETRIEVAL Presented by Sandeep...

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online