Texture Regularity Database (RegTEX)
Summary
This is a texture regularity database consisting of 21 textures exhibiting varying degrees of perceptual regularity and the their corresponding subjective regularity scores. An additional 5 sets of textures were generated from the original 21 textures through geometric and photometric transformations. These textures along with their subjective regularity scores are also available as a part of this database.
Copyright:
—— COPYRIGHT NOTICE, CONDITIONS, AND DISCLAIMER START WITH THIS LINE —————–
Copyright (c) 2014-2015 Arizona Board of Regents. All Rights Reserved.
Contact: Lina Karam ([email protected]) and Srenivas Varadarajan ([email protected])
Image, Video, and Usabilty (IVU) Lab, http://ivulab.asu.edu , Arizona State University
This copyright notice must also be included in any file or product that is derived from the database.
Redistribution and use of the texture regularity database (RegTEX), without modification, are permitted provided that the following conditions are met:
– Redistribution’s of the texture regularity database must retain the above copyright notice, this list of conditions and the following disclaimer.
– Redistribution’s in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
– The Image, Video, and Usability Laboratory (IVU Lab, http://ivulab.asu.edu) is acknowledged in any publication that reports research results using this database, copies of this database, or modifications of this database.
The papers are to be cited in the bibliography as:
S. Varadarajan and L. J. Karam, “A No-Reference Perceptual Texture Regularity Metric,” IEEE ICASSP, May 2013.
S. Varadarajan, “Texture Structure Analysis,” Ph.D. thesis, Arizona State University, May 2014.
DISCLAIMER:
This database is provided by the copyright holders and contributors “as is” and any express or implied warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed.
In no event shall the Arizona Board of Regents, Arizona State University, IVU Lab members, authors or contributors be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including, but not limited to, procurement of substitute goods or services; loss of use, data, or profits; or business interruption) however caused and on any theory of liability, whether in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of the use of this database, even if advised of the possibility of such damage.
—— COPYRIGHT NOTICE, CONDITIONS, AND DISCLAIMER END WITH THIS LINE —————–
Irregularities in Textures
The pattern of visual intensities that is spatially repeated throughout the texture in some regular or irregular manner is called a primitive. The primitives of a texture exhibit varying degrees of similarity in the visual properties like size, shape, color and orientation. These properties of the primitives along with the degree of periodicity in their placements, determine the overall perceived regularity of the primitives. The texture structure can also be quantified through the local properties of the primitives as well as through the organization amongst the primitives. Texture structure manifests itself as the perceived regularity of the primitives. Sometimes the irregularity in a texture is so high that it becomes very difficult to locate and define a primitive. Texture Structure Analysis essentially involves understanding the regularity present in a texture.
Based on the combined regularities in each of the visual properties, textures can be broadly classified as regular, hybrid and irregular textures.
Subjective testing was conducted on 21 textures. These textures were obtained from two databases, namely, the MIT Vistex database and the Graph-Cut texture synthesis database. The textures were chosen such that they distinctly represented one or more of the primitive irregularities. Also, to understand the difference in the perceived regularity between fine granular and large sized primitives, textures from both these classes were chosen in the test set. Ten subjects with normal to corrected-normal vision participated in the subjective tests. The textures, equally distributed amongst the broad classes of regular, irregular and hybrid textures, were randomly displayed one after another to each subject. The subjects were asked to score the overall regularity for each observed texture using a three-scale score with 1 corresponding to lowest and 3 to highest. In addition to observing the overall texture regularity, subjects were asked to observe and score using a three-scale score (1 being the lowest and 3 the highest) five visual properties of the texture primitives, namely (i) ease in locating the primitive, (ii)~regularity in the placement of the primitives, (iii) regularity in size, shape and color of the primitives, (iv) regularity in the direction of the primitives, and (v) average size of the primitives. The subjects also gave a final overall regularity score for each displayed texture image.
The robustness of the proposed texture regularity metric to geometric and photometric transformations was tested by applying a set of affine and color transformations. A 2D affine transformation matrix can be expressed as a decomposition of translation, rotation and scaling matrices.
Two sets of affine transformations were obtained respectively using two different parameter sets.
Photometric Transforms
Three sets of color transformations were applied by modifying the hue, saturation or contrast of the 21 textures. The hue transformed textures were obtained by transforming the colors to the CIE Lab color space and rotating the color space by 15 degrees. The saturated textures were obtained by increasing the color components in the CIE Lab color space by a factor of 2. Finally, contrast transformation was obtained by stretching the V component in the HSV color space. The values within the upper 0.5% and lower 0.5% of the V histogram were saturated to the extremal points of the range and the remaining values were stretched between them.