Method for quantitative assessment of visual quality of digital color images

DOI: 10.31673/2409-7292.2024.040004

Abstract

Digital images are one of the most common types of data circulating in computer networks. The nature of their formation and semantic content are different, but they all have significant redundancy and are therefore attractive for steganography. Color images are of particular interest. Steganographic information hiding causes the appearance of visual distortions. This requires the formation of a metric that would quantitatively assess the changes made to establish their acceptable level. Most known algorithms for assessing visual quality require the presence of a reference image, which is quite often absent. This drawback is taken into account in the work and a method for quantitative assessment of visual quality of an image is developed that does not require a reference image. It is known that the histogram of a visually perfect image is described by a normal distribution law. The main idea of ​​the proposed method is that when calculating the visual quality of a color image, the histograms of the studied and reference images are compared. Based on the differences formed, a decision is made about the visual quality of the image. The histogram of a color image is three-dimensional, therefore, for the convenience of calculating the characteristics, it is proposed to transform it into a two-dimensional representation. The verification of the operation of the proposed method on test images showed that the results obtained completely coincided with expert estimates. The results of the practical application of the proposed method in steganography are also presented. It is shown that an increase in the number of distorted pixels and an increase in multiplexing levels leads to an increase in visual quality by 10%. This is explained by the fact that a change in the least significant bits is perceived as an increase in the high-frequency component in the image. This point should be taken into account when conducting further research.

Keywords: color images-steganocontainer, visual quality metrics, test images, color space, three-dimensional histogram.