Impact of image size reducing for image quality assesment

The article describes studies of the effect of image reduction on the quantitative assessment of their quality. Image reduction refers to the proportional reduction of horizontal and vertical image resolutions in pixels. Within the framework of these studies, correlation analysis between quantitative assessments of image quality and subjective assessments of experts was performed. For the experiments, we used images from the public TID2013 database with a resolution of 512 × 384 pixels and expert estimates of their quality, as well as photographs taken with a Nikon D5000 digital camera with a resolution of 4288 × 2848 pixels. All images were reduced in 2, 4 and 8 times. For this two methods were used: bilinear interpolation and interpolation by the nearest neighbor.

22 measures were selected to evaluate image quality. Quantitative assessment of image quality was calculated in two stages. At the first stage, an array of local estimates was obtained in the vicinity of each pixel using the selected measures. At the second stage, a global quality assessment was calculated from the obtained local ones. To summarize local quality estimates, the parameters of 16 distributions of random variables were considered.

According to the results of the experiments, it was concluded that the accuracy of the quality assessment for some measures decreases with image reduction (for example, FISH, GORD, HELM, LOEN measures). BREN and SHAR measures are recommended as the best. To reduce images, it is better to use the nearest neighbor interpolation method. At the same time, the computation time of estimates is reduced on average by 4 times while reducing images by 2 times. When images are reduced by 8 times, the calculation time decreases on average by 80 times. The amount of memory required to store the reduced images is 25 times less.

1. Ponomarenko, N. Image database TID2013: Peculiarities, results and perspectives / N. Ponomarenko, L. Jin, O. Ieremeiev, V. Lukin, K. Egiazarian, J. Astola, B. Vozel, K. Chehdi, M. Carli, F. Battisti, C.-C. Jay Kuo // Signal Processing: Image Communication. – 2015. – V. 30. – P. 57–77.

2. FastStone Image Viewer – программа для просмотра изображений в Microsoft Windows [Электронный ресурс]. – URL: https://www.faststone.org/ (дата обращения 05.05.2020).

3. Gonzalez R., Woods R. Digital image processing. Technosfera // Moscow. – 2006. – 1072 p.

4. Pertuz S., Puig D., Garcia M.A. Analysis of focus measure operators for shape-from-focus. Pattern Recognition. – 2013. – № 46(5). – P. 1415–1432.

5. Starovoitov F. V. Parameters of the curve of local estimate distribution as image quality measures / Starovoitov F. V., Starovoitov V. V. // System analysis and applied information science. – 2018. – № 3. – P. 26–41.

6. Golub Y. I. Study of no-reference local image quality assessments / Y. I. Golub, F. V. Starovoitov, V. V. Starovoitov // Vestnik BrSTU. – 2019. – № 5(118): Physics, Mathematics, Computer Science. – P. 15–18.

7. Leemis, L. M. Univariate distribution relationships / L. M. Leemis, J. T. McQueston // The American Statistician. – 2008. – V. 62. – № 1. – P. 45–53.