An Analysis of Relationship Between Image Characteristics and Compression Quality for High-Resolution Satellite Images

Abstract

Prediction of compression quality of an image at a pre-encoding stage plays a vital role in any compression model because not only does it greatly reduce the associated costs, but also provides a basis for further optimization of the compression algorithm employed. The present work aims to facilitate this prediction by establishing a relationship between the inherent features of an image and its quality post compression. Since the former varies significantly for different categories of images such as nature scene, medical, remote sensing, etc., results obtained for high-resolution satellite images are largely different from similar analyses typically conducted for nature scene images. We establish the effects of gradient-based Image Activity Measure, average gray level and image contrast on Gradient Magnitude Similarity Deviation, which has recently emerged as a highly efficient Full Reference Image Quality Assessment model.