Analysis of the variability in digitised images compared to the distortion introduced by compression

Abstract

The paper evaluates the noise which is present in digitised images of very high quality and the noise/error which results when such images are compressed and then decompressed. The variations between pairs of captured images of identical material were compared and the two best pairs of images were identified. The variations between these pairs were then compared with the variations introduced by compression and decompression of those images. We found that even lossy compression can result in significantly lower variation than that between the best pairs of original images caused by imaging noise. We report the results of a qualitative questionnaire which are in good agreement with the quantitative assessment. The conclusions suggest that given the extent of noise in the imaging process the current practice of storing lossless master digitised images could be replaced by the use of more compact compressed images, arguably with no loss of quality.

Details

Creators
Sean Martin; Malcom Macleod
Institutions
Date
Keywords
digitisation; camera; scanner; digitised image; camera noise; jpeg 2000; image compression; psnr; lisbon
Publication Type
paper
License
CC BY-SA 2.0 AT
Download
337136 bytes

View This Publication