Compression of Noisy Images Taking into Account Visual Quality: A Comprehensive Study | Chapter 6 | Research Developments in Science and Technology Vol. 8

This chapter discusses the lossy compression of pictures that include additive noise damage. The examination is done from the standpoint of compressed picture visual quality, which is the chapter's major contribution. The compression ratio for a number of coders is investigated and is controlled in various ways. Lossless coding typically doesn't yield compression ratios high enough for many real-world uses. The visual quality metrics WSNR, MS-SSIM, PSNR-HVS-M, and PSNR-HVS are employed since they are the best suitable for the application under consideration. The goal is to determine the optimal operating point (OOP) based on measurements for visual quality. It is shown that, under some circumstances, picture filtering caused by lossy compression, or the possibility of OOP, may cause compressed images to have slightly higher visual quality than the original noisy images. The "optimal" coder settings that allow for the observation of this beneficial impact depend on the noise standard deviation. In OOP, we suggest a technique for determining coder parameters. This makes it possible to create automatic methods for compressing noisy pictures close to the ideal operating point, or if visual quality increases or decreases inadvertently. Another benefit is that this case's compression ratio is fairly high. The outcomes of a collection of test photos in grayscale with different noise variations are displayed.

Author(s) Details:

Vladimir Lukin,

National Aerospace University, Kharkov, Ukraine.

Bogdan Kovalenko,

National Aerospace University, Kharkov, Ukraine.

Sergii Kryvenko,

National Aerospace University, Kharkov, Ukraine.

Nikolay Ponomarenko,

National Aerospace University, Kharkov, Ukraine and Tampere University, Tampere, Finland.

Karen Egiazarian,

Tampere University, Tampere, Finland.

Jaakko Astola,

Tampere University, Tampere, Finland.

Please see the link here: https://stm.bookpi.org/RDST-V8/article/view/7399