Widespread Dilute Smoke Correction in the Multispectral Images through Iteratively Applying the Regression Estimated Residuals to Spectral Bands
|
Reza Qalavand , Alireza Safdarinezhad * , Behzad Behnabian |
Tafresh University |
|
Abstract: (942 Views) |
Widespread wildfires and the resulting smoke are a common occurrence in various regions of the earth’s surface. Identification of the contaminated areas and crisis management requires the classification of areas affected by smoke. On the other hand, visible spectral bands in satellite images are affected by such phenomena, which blur the clarity of these images. To address this issue, a repetitive regression-based method is proposed in order to reduce the impact of dilute smoke on multispectral images. The occurrence of large residuals when the affected bands are estimated with the non-affected bands is the motivation for designing this method. The method involves iteratively applying residuals from regression models to probable smoke areas after refinement and localization, which corrects spectral observations in affected bands. This solution maintains radiometric content in clean image areas and significantly improves image clarity in smoke-contaminated areas. The results show quantitative improvement in the correlation between corrected images and smoke-free images in the most affected spectral bands, averaging 14.2 percent. This method can only be used for dilute smoke with a visible background.
|
|
Keywords: Linear Regression, Dilute smokes, Multispectral Images, Scattering of light, Atmosphere, Residuals Vector |
|
Full-Text [PDF 2117 kb]
(298 Downloads)
|
Type of Study: Research |
Subject:
RS Received: 2023/06/13 | Accepted: 2023/09/20 | ePublished ahead of print: 2024/02/18 | Published: 2024/03/4
|
|
|
|
|
Send email to the article author |
|