Satellite images are usually degraded due to inaccuracy or limitations of the transmission and storage devices. Many researchers are still trying to remove noise from satellite images. Satellite image denoising techniques should not distort edges in an image. FFDNet is an image denoising solution that was recently developed. FFDNet has several advantages over other existing neural network denoisers, including a faster execution time and reduced memory footprint, as well as the ability to properly handle a wide variety of noise levels with a single network model. This technique is appealing for practical denoising applications because of its denoising performance and minimal computing load. High-Resolution LAPAN-A3 imagery has great potential for use in various remote sensing applications. But it has a problem related to its dynamic range which causes very high noise that will appear when we try to increase the brightness of the image. In this paper a different approach for LAPAN-A3 satellite imagery denoising, namely BM3D, FastNLM, and FFDNet. This method has tried to compare in terms of denoising performance, with three different cases of AWGN, model, and mixed noise. The result shows that FFDNet has the best performance among others, especially when recovering images from AWGN noise. But when trying to remove the model and mixed noise, all of these methods have almost the same output value, but FFDNet is slightly better than others.
Denoising LAPAN A3 High-Resolution Digital Camera Images: A Comparative Study
2021-11-03
1649619 byte
Conference paper
Electronic Resource
English
Lidar activity at Lapan, Indonesia
IEEE | 1999
|Development of camera lens focus controller on LAPAN-A4 satellite payload
American Institute of Physics | 2021
|