VirtualDub MSU Noise Estimation Filter
Project, idea: Dr. Dmitriy Vatolin, Sergey Grishin
Algorithm, implementation: Kumok Boris
Updating and additions: Sheludko Victor
This filter is used to estimate mean noise variance in video sequences. The result is saved into a log file.
The basic menu of the filter
Noise estimation algorithm - the algorithm used to estimate noise. Following modes supported:
- Spatio-temporal gradients
- All algorithms
Advanced - configure algorithm details
Write noise estimates for each frame into file - path to log file.
Configure algorithm details: Block-Based
Block size - The parameter influencing on speed of work and accuracy of an estimation. The increase of this parameter reduces accuracy and reduces time of performance. For video with small homogeneous areas it is recommended to use smaller values.
Search step - The parameter influencing on speed of work and stability of an estimation. Great values increase speed of work and reduce stability of an estimation.
Rate - The parameter influencing on reliability and stability of an estimation. Great values are reasonable for using for video with the big homogeneous areas.
Margin - The size of unused edges of the image.
Brightness border - The parameter influencing reliability of an estimation. For video visually strongly polluted it is recommended to increase this parameter.
Configure algorithm details: Spatio-Temporal Gradients
Evaluation stability - The parameter influencing on stability of an estimation. The increase of this parameter increases stability, but reduces accuracy of an estimation.
Brightness border - The parameter influencing on reliability of an estimation. For video visually strongly polluted it is recommended to increase this parameter.
Job control & AviSynth
The filter supports Job Control, which allows to use it in AviSynth. Example:
LoadVirtualDubPlugin("...\VirtualDub\plugins\MSUNoiseEstimator.vdf","MSU_Noise_Estimator", 0) clip=AVISource("...\clip_input.avi", false, "RGB24") clip.ConvertToRGB32.MSU_Noise_Estimator("C:\log.csv", 0)
|0||File||Path to log file|
0 - MAD
1 - Block-Based
2 - Spatio-Temporal Gradients
3 - All algorithms
Each graph shows the examples of noise detection on 4 video sequences: the original one and 3 with noise added.
"Shattered" test sequence. Noise was added using MSU Noise Generator v2.1.
"Susie" test sequence with the removed odd frames. Noise was added using MSU Noise Generator v2.1.
|For common questions about filter's work please contact us|
|For commercial license of this filter (commercial usage is not free) please contact us|
|Last updated: 17-January-2020|
Project updated by
Server Team and MSU Video Group
Project sponsored by YUVsoft Corp.
Project supported by MSU Graphics & Media Lab