Visual Content Indexing and Retrieval with Psycho-Visual Models by Jenny Benois-Pineau & Patrick Le Callet

Author:Jenny Benois-Pineau & Patrick Le Callet
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

(1)

An image that depicts all the requested shapes with the same relative projected area would have the highest entropy. Since the relative projected areas form a probability distribution, the relative visibility of the background at maximum entropy () should be also comparable to the visibility of each shape. In practice, we do not reach this upper bound and simply maximize E(I j ) over j.

We return the image with the highest entropy as the result of our query. The system then switches the video stream to the one corresponding to the resulting image.

As clearly stated by Vazquez et al. the intervention of helps to handle various zoom levels (or various distances between the cameras and the scene) among the J candidate images. The use of the background visibility level gives nearer shapes a higher entropy. In Fig. 3, a larger projection of the requested yellow shape increases the entropy of the best viewpoint (by limiting the information brought by the background).

Fig. 3A 3D query volume V q intersecting K q = 2 shapes and J = 2 candidate viewpoints

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.