10 October 2017 Output MSE and PSNR prediction in DCT-based lossy compression of remote sensing images
Author Affiliations +
Amount and size of remote sensing (RS) images acquired by modern systems are so large that data have to be compressed in order to transfer, save and disseminate them. Lossy compression becomes more popular for aforementioned situations. But lossy compression has to be applied carefully with providing acceptable level of introduced distortions not to lose valuable information contained in data. Then introduced losses have to be controlled and predicted and this is problematic for many coders. In this paper, we analyze possibilities of predicting mean square error or, equivalently, PSNR for coders based on discrete cosine transform (DCT) applied either for compressing singlechannel RS images or multichannel data in component-wise manner. The proposed approach is based on direct dependence between distortions introduced due to DCT coefficient quantization and losses in compressed data. One more innovation deals with possibility to employ a limited number (percentage) of blocks for which DCT-coefficients have to be calculated. This accelerates prediction and makes it considerably faster than compression itself. There are two other advantages of the proposed approach. First, it is applicable for both uniform and non-uniform quantization of DCT coefficients. Second, the approach is quite general since it works for several analyzed DCT-based coders. The simulation results are obtained for standard test images and then verified for real-life RS data.
© (2017) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Ruslan A. Kozhemiakin, Ruslan A. Kozhemiakin, Sergey K. Abramov, Sergey K. Abramov, Vladimir V. Lukin, Vladimir V. Lukin, Benoit Vozel, Benoit Vozel, Kacem Chehdi, Kacem Chehdi, } "Output MSE and PSNR prediction in DCT-based lossy compression of remote sensing images", Proc. SPIE 10427, Image and Signal Processing for Remote Sensing XXIII, 1042721 (10 October 2017); doi: 10.1117/12.2278002; https://doi.org/10.1117/12.2278002

Back to Top