Perceptual similarity between color images using fuzzy metrics

Abstract

In many applications of the computer vision field measuring the similarity between (color) images is of paramount importance. However, the commonly used pixelwise similarity measures such as Mean Absolute Error, Peak Signal to Noise Ratio, Mean Squared Error or Normalized Color Difference do not match well with perceptual similarity. Recently, it has been proposed a method for gray-scale image similarity that correlates quite well with the perceptual similarity and it has been extended to color images. In this paper we use the basic ideas in this recent work to propose an alternative method based on fuzzy metrics for perceptual color image similarity. Experimental results employing a survey of observations show that the global performance of our proposal is competitive with best state of the art methods and that it shows some advantages in performance for images with low correlation among some image channels.

Fuzzy partial metric spaces

“…Since both concepts of fuzzy metrics were defined, several authors have contributed to their study from the mathematical point of view (see, for instance, the following current references Gutiérrez-Gracía et al (2018); Miñana and Valero (2018); Shukla et al (2016), and references therein). Moreover, fuzzy metrics have been used successfully in engineering applications such as colour image filtering (see Camarena et al (2010), and references therein) and perceptual colour difference (see Grečova and Morillas (2016); Gregori et al (2012)).…”

Colorizing the Past: Deep Learning for the Automatic Colorization of Historical Aerial Images

“…Color difference evaluation is a complex and investigated task [ 67 , 68 , 69 , 70 ]. Studies in this field aim to identify a comprehensive formulation for objectively quantizing color differences, considering the influence of many factors on color perception and comparison.…”

Using suprathreshold color-difference ellipsoids to estimate any perceptual color-difference

“…Improved correlation between visually perceived (∆V ) and instrumentally measured (∆E) colour differences under specific illuminating and viewing conditions is an important problem in modern colorimetry. This topic is gaining more and more attention each day in the computer vision field as many recent image processing and computer vision techniques are using colour difference formulas when addressing perceptual processing and understanding of digital images [1]- [3]. For instance, advanced models of perceptual similarity of color images iCAM [1] or S-CIELAB [2] try to represent the Human Visual Systems mechanisms and are based on using appropriate Constrast Sensitivity Functions (CSFs) to remove all details in the images that cannot be perceived as well as Color-difference formulas to perceptually characterize the differences which are indeed observed, which has been found to be appropriate in general.…”

“…For instance, advanced models of perceptual similarity of color images iCAM [1] or S-CIELAB [2] try to represent the Human Visual Systems mechanisms and are based on using appropriate Constrast Sensitivity Functions (CSFs) to remove all details in the images that cannot be perceived as well as Color-difference formulas to perceptually characterize the differences which are indeed observed, which has been found to be appropriate in general. In turn, perceptual similarity measures assess how good are filtering methods, compression algorithms or demosicing methods from a perceptual point of view [3].…”