Prediction of optimal operation point existence and parameters in lossy compression of noisy images

Abstract

This paper deals with lossy compression of images corrupted by additive white Gaussian noise. For such images, compression can be characterized by existence of optimal operation point (OOP). In OOP, MSE or other metric derived between compressed and noise-free image might have optimum, i.e., maximal noise removal effect takes place. If OOP exists, then it is reasonable to compress an image in its neighbourhood. If no, more "careful" compression is reasonable. In this paper, we demonstrate that existence of OOP can be predicted based on very simple and fast analysis of discrete cosine transform (DCT) statistics in 8x8 blocks. Moreover, OOP can be predicted not only for conventional metrics as MSE or PSNR but also for visual quality metrics. Such prediction can be useful in automatic compression of multi-and hyperspectral remote sensing images.

Lossy compression of noisy remote sensing images with prediction of optimal operation point existence and parameters

“…In Ref. 29, 12 grayscale test images have been used. Eight of them can be considered as the standard optical test images: Baboon, Stream&Bridge, Man, Lenna, Tiffany, F − 16, Peppers, Boats.…”

“…Keeping in mind similarities between denoising and the lossy compression of noisy images, a similar approach has been proposed to predict the OOP existence and its compression parameters in our paper. 29 The advantage of this approach is its high computational efficiency. It uses DCT in 8 × 8 pixel blocks that is a standard operation in image processing.…”

“…Our idea is the following. 28,29 If OOP exists for a given (sub-band) image, it is worth carrying out compression in an OOP neighborhood. If OOP does not exist, it is worth performing a more careful compression (introducing fewer losses).…”

“…Our recent paper 29 demonstrated that there is an approach that works rather well in predicting OOP existence and the corresponding compression parameters for lossy noisy image compression using four coders. There are several aspects not considered in this paper 29 and the obtained results have some shortcomings. First, a limited number of test images (12) and noise standard deviation values (3) have been used to obtain prediction dependences by curve fitting.…”

“…Thus, the present paper can be regarded as an extended version of Ref. 29, where the aforementioned aspects are paid special attention. Its main novelty consists in a specific adaptation of PCC to image/noise properties in hyperspectral data sub-bands.…”