A MATLAB implementation Psychology and Art Theory-based Features (MH)
This set of MATLAB scripts extract the MH features [1] used in [2]. This code does NOT run itself out-of-the-box since it depends on many other programs and the AIC dataset. While we tried our best to make the job as simplified as possible, some of them must be downloaded and compiled separately due to license/compatibility issues. The code to compute Tamura texture features were ported from a C code of [7] under the MIT license included.
You must cite [1] in order to use these features. In addition, please consider citing [2] if you found this code useful.
Please let me know if you found any bug(s).
STEP 1. Download hsy_rgb.zip from http://allan.hanbury.eu/. The file is located under Colour Resources (as of 2015/02/10). Copy the two files into the same directory of this file: hsy2rgb.m, rgb2hsy.m
STEP 2. Download FastEMD solver from [3]. Compile. Copy following two files into the same directory of this file: emd_hat_gd_metric_mex.mex, emd_hat_mex.mex
STEP 3. Download Color Descriptor from [4]. Compile. Copy following files into the same directory of this file: mexColorNaming.mex*, w2c.mat
STEP 4. Download color space converter from [5]. Compile. Copy following files into the same directory of this file: colorcalc.mex*, colorspace.m
STEP 5. Download RGB histogram calculator from [6]. Copy following file into the same directory of this file: rgbhist_fast.m
STEP 6. You need the MATLAB Computer Vision Toolbox. If not, you need to obtain a Viola-Johns Face Detector implementation somewhere.
STEP 7. You need to obtain AIC dataset to properly train the Fuzzy C-Means. While I included pretrained models (AICfcm.mat), it is strongly recommended to train them again. Particularly, if you were to reproduce [1], you must re-train it using IAPS dataset instead of the pretrained models. Please check out the codes fuzzyFuncLearn.m for more details.
A MATLAB implementation Psychology and Art Theory-based Features (MH)
Copyright (C) 2014 Sejong Yoon (sjyoon@cs.rutgers.edu)
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation, version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program. If not, see <http://www.gnu.org/licenses/>.[1] J. Machajdik and A. Hanbury, Affective Image Classification using Features Inspired by Psychology and Art Theory, ACM Intl. Conf. on Multimedia (MM), 2010. URL: http://www.imageemotion.org/ [2] S. Yoon and V. Pavlovic, Sentiment Flow for Video Interestingness Prediction, ACM Intl. Conf. on Multimedia (MM) Workshop (HuEvent), 2014. [3] O. Pele and M. Werman, Fast and Robust Earth Mover's Distances, ICCV 2009. URL: http://www.ariel.ac.il/sites/ofirpele/FastEMD/ [4] J. van de Weijer, C. Schmid, J. Verbeek, D. Larlus Learning Color Names for Real-World Applications, IEEE Trans. in Img. Proc. (TIP), vol 18 (7):1512-1524, 2009. URL: http://cat.uab.es/~joost [5] P. Getreuer MATLAB Central File Exchange - Colorspace Transformations URL: .../28790-colorspace-transformations [6] M. K. Reddy, MATLAB Central File Exchange - Color Histogram of an RGB Image. URL: .../43630-color-histogram-of-an-rgb-image [7] T. Minka and R. W. Picard, A Sample Implementation of Tamura Texture Feature in C. URL: http://vismod.media.mit.edu/pub/tpminka/features/
Last Updated: February 10, 2015