The Importance of Phase to Texture Similarity

Xinghui Dong, Ying Gao, Junyu Dong, Mike J. Chantler; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 2758-2766

Abstract


Although the importance of the Fourier phase to image perception has been addressed, it is unknown this is the case for texture similarity or not. We first show that phase is more important to human perceptual texture similarity than magnitude. We further test the ability of 51 feature sets to use phase for texture similarity. Yet it is found that magnitude is more important to these feature sets than phase. Considering the inconsistency between the similarity data obtained using humans and those feature sets, we attribute this to the difference in the ability of humans and these feature sets to use phase. Thus, we enable the 51 feature sets to use phase by fusing the features extracted from the original and phase-only images. It is shown that the fused feature sets yield better results than those derived using the 51 feature sets. In particular, this finding can also be propagated to CNN features. These promising results should be due to the importance of phase to texture similarity.

Related Material


[pdf]
[bibtex]
@InProceedings{Dong_2017_ICCV,
author = {Dong, Xinghui and Gao, Ying and Dong, Junyu and Chantler, Mike J.},
title = {The Importance of Phase to Texture Similarity},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2017}
}