Please note that the publication lists from Infoscience integrated into the EPFL website, lab or people pages are frozen following the launch of the new version of platform. The owners of these pages are invited to recreate their publication list from Infoscience. For any assistance, please consult the Infoscience help or contact support.
Joint Statistical Analysis of Images and Keywords with Applications in Semantic Image Enhancement
With the advent of social image-sharing communities, millions of images with associated semantic tags are now available online for free and allow us to exploit this abundant data in new ways. We present a fast non-parametric statistical framework designed to analyze a large data corpus of images and semantic tag pairs and find correspondences between image characteristics and semantic concepts. We learn the relevance of different image characteristics for thousands of keywords from one million annotated images. We demonstrate the framework’s effectiveness with three different examples of semantic image enhancement: we adapt the gray-level tone-mapping, emphasize semantically relevant colors, and perform a defocus magnification for an image based on its semantic context. The performance of our algorithms is validated with psychophysical experiments.
Proceedings of the 20th ACM international conference on Multimedia (MM’12)
2012
ACM Multimedia, Osaka, Japan, October 29 – November 2, 2012.p. 489-498
DOI : 10.1145/2393347.2393417
Example Images
Semantic tone-mapping:
input |
|
output: dark |
output: sand |
See more example images for semantic tone-mapping and the psychophysical experiments.
Color enhancement:
input |
output: strawberry |
See more example images for color re-rendering.
Depth-of-field adaptation:
input |
output: macro |
See more example images for spatial frequency re-rendering.
Supplementary material for the statistical framework
- In this publication we use the Mann-Whitney-Wilcoxon (MWW) test to determine the significance values z. There are of course other tests that also estimate whether two distributions are significantly different from each other. An overview is given here, together with an explanation why we favor the MWW test.
- Figure 5 in the article shows the Δz* values for 50 selected keywords. Here is a complete overview for all 2858 keywors and all descriptors.
Download
Download Matlab package with pre-computed significance values (40 MByte).
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.
Contact
Questions or feedback are welcome. Please mail the author or visit his website.