User:KatyaShm/sandbox

for PSNR:

Application
PSNR has applications in a variety of different problems. Some examples are:


 * Image compression. PSNR is commonly used to quantify reconstruction quality for images and video subject to lossy compression.
 * Image restoration. PSNR is one of the most popular metrics for the quality assessment of the restored images.
 * Steganography. PSNR is often used to evaluate the steganography methods quality.
 * Improvement of fabricated machines. PSNR criterion is employed to analyze the fabricated machine performance which aids its improvement. For example, the PSNR criterion is used to improvement of an apple sorting machine.
 * Image segmentation. PSNR is used as an analytic metric by several authors of threshold-based segmentation algorithms.
 * 3D video. PSNR used as estimation Method for 3D videos.
 * Super-resolution imaging. PSNR is the most popular metric for the quality assessment of Super-Resolution algorithms.

Criticism
PSNR has been widely criticized in independent research papers.
 * There are many examples when images with mostly the same PSNR scores have dramatically different visual quality.
 * In the task of representing the quality of a compressed video sequence, PSNR has been found to correlate poorly with subjective quality ratings.
 * PSNR is very sensitive to small geometric transformations, which doesn't alter visual quality a lot.
 * PSNR is not an adequate quality measurement for segmentation algorithms.
 * PSNR doesn’t reflect the actual quality of stego images.
 * PSNR correlates poorly with subjective assessment for the task of Super-Resolution.
 * Ways of cheating on popular objective metrics: blurring, noise, super-resolution, and others [].

for SSIM:

Application

 * Image denoising. SSIM is used as a metric for the quality assessment of denoised images.
 * Loss function. SSIM loss is one of the most widely used loss functions in training neural networks for image processing tasks.
 * Super-resolution imaging. SSIM is the most popular metric for the quality assessment of Super-Resolution algorithms.

Criticism
SSIM has been widely criticized in independent research papers.
 * Jim Nilsson and Tomas Akenine-Möller questioned the philosophy of perceptual motivated nature that was laid in the SSIM.
 * SSIM is very sensitive to small geometric transformations, which doesn't alter visual quality a lot.
 * Keyan Ding, Kede Ma, Shiqi Wang, and Eero P. Simoncelli evaluated SSIM and other full-reference metrics by using them as objective functions to train deep neural networks for four low-level vision tasks: denoising, deblurring, super-resolution, and compression.
 * SSIM has been found to correlate poorly with subjective assessment in the task of representing the quality of Super-Resolution images.
 * Ways of cheating on popular objective metrics: blurring, noise, super-resolution, and others [].