Premium
Psychophysical Validation of A Digital Method to Assess Ill‐Defined Visual Boundaries: An Example with Fabric Openness Factor
Author(s) -
Tàpias Montserrat,
Ralló Miquel,
Escofet Jaume
Publication year - 2015
Publication title -
journal of sensory studies
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.61
H-Index - 53
eISSN - 1745-459X
pISSN - 0887-8250
DOI - 10.1111/joss.12186
Subject(s) - computer science , digital image , artificial intelligence , moment (physics) , visual inspection , process (computing) , openness to experience , digital image analysis , computer vision , feature (linguistics) , pattern recognition (psychology) , image (mathematics) , statistics , mathematics , algorithm , image processing , psychology , social psychology , linguistics , philosophy , physics , classical mechanics , operating system
Fabric openness factor ( OF ) is the fraction of the web area that is uncovered by yarns. OF is a critical feature regarding the end‐use performance of the fabric and should be accurately assessed. However, digital OF estimates yielded by image binarization algorithms differ among them depending on the criteria used, mainly due to ill‐defined boundaries, thus precluding a straightforward assessment of the actual fabric OF value. Lacking any standard to compare actual OF values with measured OF values, we addressed the validation procedure of the digital assessment method from visual OF estimates. OF of 81 distinct fabric samples was evaluated from digital images by a panel of 18 observers using visual binarization technique. Following the psychophysical models of Fechner and Stevens, these visual estimates were correlated with digital estimates yielded by several binarization algorithms. Stevens’ psychophysical model and an automatic binarization algorithm developed by us scored the highest correlation. Practical Applications This work addresses the problems that arise at the moment of measuring the area of a region with ill‐defined boundaries without any standard to control the process of measurement. Similar problems are also found in important fields such as medical imaging. In these cases, there are usually different digital methods that give rise to estimates of the magnitude which mismatch among them. To solve these differences, a more common approach is to use estimates provided by a panel of observers and consider them to be objective, regardless of the bias between subjective responses and actual values. However, visual estimates do not show either accuracy (because they are psychophysical measures) or precision (due to the variability among observers). To overcome these difficulties, we propose a procedure to validate a digital method through visual estimates provided by a panel of observers, taking into account the psychophysical models of Fechner and Stevens.