Summary
International Workshop on Smart Info-Media Systems in Asia
2022
Session Number:RS2
Session:
Number:RS2-2
Low-Level Gaussian Noise Estimation by Using the Intercept of the Regression Line Derived from Standard Deviation of Multiple Blocks of Divided Images
Takashi Suzuki, Hiroyuki Tsuji, Tomoaki Kimura,
pp.101-106
Publication Date:2022/9/15
Online ISSN:2188-5079
PDF download
Summary:
A robust estimation method based on MAD (Median Absolute Deviation) can be used to estimate the standard deviation of the Gaussian noise superimposed on the image. The estimation method can be used with good accuracy if the standard deviation of the Gaussian noise superimposed on the image is greater than 10. However, if the standard deviation is less than 10 and the image contains many edges and details, it becomes difficult to separate the edge and detail signals from the superimposed noise, resulting in not good estimation accuracy. Therefore, in this paper, we propose a noise estimation method that finds a regression line from multiple blocks in images with many edges and detail signals, and uses the intercept as the Gaussian noise to be estimated. The proposed method is found to perform well for images with many edges and detail signals.