MSU Noise Estimation Metric (NE)

Common idea

Metrics Noise Estimator is intended for calculation of noise level for each frame of video sequence.


[!] — Known bug
[+] — New Feature
[*] — Other

[*] Incorrect processing of non-standard resolution video bug fixed

[*] Bug with incorrect (identical) values for some videos is fixed.
[*] Home page is fixed (was incorrect)
[*] Command line name is changed (became easier to use)

[*] First released version.


The metrics realizes three various algorithms of definition of noise level:

The choice of which algorithm to use can be made in Settings. To algorithms there correspond figures from 0 up to 2.


Visualization of the metrics does not carry any information.


By results of job of the metrics the plot of frame-accurate value of noise level is constructed. Final value of the metrics is average arithmetic of all frame-accurate values.

Plot's example
Plot's example



For each frame do HAAR wavelet decomposition. Than evaluate median of HH-component’s absolute values. Final value of the metrics is the normalized median.


Frames are tessellated into a number of 8x8 blocks. Standard deviations of intensity (measures of intensity variation) are computed for all the blocks and sorted. The block with the smallest standard deviation has the least change of intensity. The smaller the standard deviation, the smother the block. The intensity variation of a smooth block may be due to noise, in which the standard deviation of the block is close to that of the Gaussian noise added. Normalized average arithmetic values of 30 % of all blocks with the least values grows is the final value of the metric.

Spatio-Temporal Gradients

For each frame is doing wavelet decomposition. computing temporal and spatial histograms. The initial estimation of noise level is defined by value at which temporal or spatial histogram achieves the maximal value. The decision of whether to use the spatial or temporal histogram is based on the deviation of the histogram from the Rayleigh distribution. Then this estimation is corrected, using Kolmogorov-Smirnoff test. The normalized corrected estimation is the final value of the metric.




10 May 2017
See Also
Parallax range estimation in S3D video
Parallax determines the depth of S3D movies. The range of parallaxes should be both comfortable and entertaining for spectators.
MSU Video Quality Measurement Tool: Why upgrade?
MSU Video Quality Measurement Tool (PSNR, MSE, VQM, SSIM)
MSU Video Quality Measurement Tool (VQMT) is a program for objective video quality assessment.
MSU VQMT Subjective quality correlation
Comparison provided for the most popular metrics: PSNR, SSIM, 3-SSIM, MS-SSIM and new stSSIM
MSU Video Quality Measurement Tool: SDK
MSU Perceptual Video Quality: Subjective video quality methods information
Site structure