Abstract
We address lossy compression of noisy remote sensing images, where the noise is supposed to be spatially uncorrelated (white), additive originally or after a proper variance-stabilizing transformation (VST). In such situations, the so-called optimal operation point (OOP) might exist. The OOP is associated with the parameter that controls compression (e.g., quantization step) for which the compressed image is the closest to the noise-free image according to a certain criterion and is closer than the original noisy image. Lossy compression in the neighborhood of OOP, if it exists, relates to an essential noise filtering effect and some distortions. Then such lossy compression (in the neighborhood of OOP) becomes expedient. However, it may be that OOP does not exist for a given image and the observed noise intensity. In such a situation, it can be reasonable to carry out a more “careful” image compression (with a lower compression ratio). Also, it is expedient to predict the existence of OOP and the compression parameters at this point in advance in order to perform adaptive and automated compression. The OOP existence that can be predicted for some coders based on the discrete cosine transform (DCT) is shown. The proposed prediction procedure is simple and fast. It presumes the calculation of DCT coefficient statistics in nonoverlapping 8×8 pixel blocks for a given image and uses an approximating curve obtained in advance. It is shown that it is possible to predict values for both conventional metrics, such as mean square error or peak-signal-to-noise ratio, and some visual quality metrics for the coder parameters that correspond to a possible OOP. The designed prediction procedure is tested on Hyperion and AVIRIS hyperspectral remote sensing data.
Highlights
Today, Earth monitoring systems find more and more applications.[1,2] Most modern remote sensing (RS) systems are multichannel
This means that the advanced DCT (ADCT) coder provides a slightly better visual quality of the compressed noisy images
We have demonstrated that the existence of operation point (OOP) for a given image corrupted by AWGN can be quite accurately predicted if the variance of AWGN is a priori known or accurately pre-estimated
Summary
Earth monitoring systems find more and more applications.[1,2] Most modern remote sensing (RS) systems are multichannel (e.g., full-polarization, multispectral or hyperspectral). The same observations can be found for the coder AGU applied to other images and noise variance values in the paper.[20] The positions (arguments) of OOP slightly differ depending upon an image and noise variance, but QSOOP ≈ ð3.5: : : 4.0Þσ0 can be considered as the recommended value to attain OOP (if it exists) for the AGU coder.[20] This means that the compression in OOP or its close neighborhood can be provided without having a noise-free image if σ0 is known in advance or accurately pre-estimated. One standard metric frequently used in lossy compression analysis is MSE, defined as
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
