MI07slides - . . Subject 07: Image Retrieval I...

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon
. . Subject 07: Image Retrieval I Content-Based Image Retrieval Retrieval is a process of recalling information. Images may be located by textual descriptors, aiming at describing their content. For example, “sunrise Easter Island” lead on 1 March 2008 to 91,700 hits on Google’s image search engine, starting with this page: About 50% on this page is related to a sunrise on Easter Island. Textual descriptors of image content are doomed to failure (“A picture is worth a thousand words.”). Page 1 July 2009
Background image of page 1

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

View Full DocumentRight Arrow Icon
. . Subject 07: Image Retrieval I Content-Based Visual Information How to describe the differences between those images (taken at Tongariki and Rano Roraku on 14 November 2007)? Instead of words, we may use visual information in pictures, such as colors, shapes of objects, subwindows, a copy at lower resolution, and so forth, for image search. Those descriptors can be given in pictorial or other form. Page 2 July 2009
Background image of page 2
. . Subject 07: Image Retrieval I Descriptors might be, for example, curves (frontiers of image regions), rectangular or elliptical subwindows (at a fixed scale, “visual nouns”, considering a picture to be a “visual word”), separating curves (such as horizon), histograms, counts of particular objects or features (number of moais, of mountain peaks, and so forth). Histograms are a simple way of characterizing value distributions, and texture characterization is the general approach. Page 3 July 2009
Background image of page 3

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

View Full DocumentRight Arrow Icon
. Subject 07: Image Retrieval I If possible, local descriptors (i.e., for windows of some constant size) are preferred compared to global descriptors (possibly based on analyzing the whole picture, without any a-priori size limitation), for reasons of simplicity or time-efficiency. The following synthesized pictures illustrate the difficulty of understanding global patterns (such as disk, or halfplane) if only local information is available, or if pictures are given at different scales. (from J. Richardt, F. Karl, R. Klette, and C. M¨uller, 1996)
Background image of page 4
Image of page 5
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 11

MI07slides - . . Subject 07: Image Retrieval I...

This preview shows document pages 1 - 5. Sign up to view the full document.

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