JPEG compression is an image coding method for saving storage/bandwidth capacity while maintaining high perceived quality. Although the JPEG recommendation (ITU-T T.81 | ISO/IEC 10918-1) dates back to 1992/1993, JPEG is still one of the most popular image coding formats today. JPEG compression achieves much of its storage efficiency by transforming the spatial image into the frequency domain and quantizing individual spectral frequencies based on the perceived visibility to the human eye. In principle, the quantization factors can be chosen freely. However, the example quantization tables provided in Annex K of the JPEG recommendation have become the de facto standard, probably due to their adoption in the popular JPEG implementation libjpeg. As a result, these quantization tables play a significant role in academic research on image forensics and steganalysis.
The goal of this Bachelor’s thesis is to replicate the psychovisual experiments that lead to the quantization tables provided in Annex K. This involves creating stimuli images which show individual DCT basis functions with increasing visibility. The student will then design and run a perceptual study where participants are shown these stimuli images and determine the visibility threshold when they can differentiate the basis function from the background. Ideally, the study will compare the results between an old cathode ray tube (CRT) and a modern LCD display.