RUS  ENG
Full version
JOURNALS // Vestnik of Astrakhan State Technical University. Series: Management, Computer Sciences and Informatics // Archive

Vestn. Astrakhan State Technical Univ. Ser. Management, Computer Sciences and Informatics, 2021 Number 4, Pages 58–67 (Mi vagtu693)

This article is cited in 2 papers

COMPUTER SOFTWARE AND COMPUTING EQUIPMENT

Algorithm and methods of ranking group of bitmap images

V. V. Afonin, A. V. Savkina, V. V. Nikulin

National Research Ogarev Mordovia State University, Saransk, Republic of Mordovia, Russian Federation

Abstract: The article presents an algorithm and a methodology of ranking a group of raster images by using the criterion of their expected quality. Ranking refers to the evaluation of a sample of bitmap images in a descending order of their quality, the image quality assessment being performed on the basis of a number of statistical parameters, such as coefficients of variation, determination, rank correlation index, as well as errors (absolute maximum error, average error, average quadratic error). The differences between the images are based on converting a full-color RGB image into HSV, Lab, NTSC, XYZ, YCbCr color models, which are represented as one-dimensional pixel matrices. The colour model RGB is taken as a reference. In relation to it, the proposed statistical characteristics of other color models are compared, any object of each color model being compared with the base model - an RGB image. Based on this comparison, all images of a given group are analyzed independently of each other. Image quality assessment is performed in a module that can be used to cycle through multiple images and is represented in numerical form as a real number. One of the module blocks calculates the statistical parameters between each color model and the base RGB model. After receiving the values of the quality scores they are ranked according to their values. As a result, an image with a higher or lower scene quality can be determined. Images with blocking artifacts, noisy images of the salt & pepper type, and images with strobe effects artifacts were considered as test images.

Keywords: color spaces, colour models, bitmap images, coefficients of variation, determinations, rank correlation, maximum, average, mean-square errors.

UDC: 004.932.2

Received: 30.07.2021

DOI: 10.24143/2072-9502-2021-4-58-67



© Steklov Math. Inst. of RAS, 2024