Open AccessNoise Level Based Denoising Technique Utilizing Patch- Based Noise Level Estimator for Low-Light Condition Surveillance Image
Author(s) -
Suhaila Sari,
Wong Zhi Lin,
Hazli Roslan,
Nik Shahidah Afifi Mohd Taujuddin,
Chua King Lee,
Siti Zarina Mohd Muji
Publication year - 2018
Publication title -
international journal of engineering and technology
Language(s) - English
Resource type - Journals
ISSN - 2227-524X
DOI - 10.14419/ijet.v7i4.30.28173
Subject(s) - noise reduction , noise (video) , artificial intelligence , computer science , median filter , computer vision , dark frame subtraction , filter (signal processing) , image noise , noise measurement , salt and pepper noise , value noise , otsu's method , pattern recognition (psychology) , image processing , image (mathematics) , image segmentation , noise floor
Digital Image Processing is a method to obtain image or to take out useful details or feature from image. The noise will cause the results of error in the image acquisition process. Generation of higher noise levels in the low light condition environment will often result in oversmoothed edges and textures during the denoising process because of lower signal levels in the image. Thus, this study goal is to improve denoising techniques for Poisson noise removal in low light condition for surveillance images. The Patch-Based Noise Level Estimator is designed to estimate the noise level of noisy image. The noisy image then fed to either OTSU WIE-WATH Filter or OTSU KU-WIE-WATH Filter automatically based on the noise level of image. The OTSU WIE-WATH Filter is used for low and medium Poisson noise removal while OTSU KU-WIE-WATH Filter is used mainly for high Poisson noise removal. The proposed denoising technique performances are analyzed with other existing denoising techniques in terms of Mean Absolute Error (MAE), computational time and visual effect inspection. The results verified that proposed technique is effective in removing different level Poisson noise in low light condition surveillance images.