Human Perception Oriented Image Enhancement

Abstract

We developed a DCT domain image enhancement method that not only improves human perceived quality, but also overcomes artifacts that inherently occur from image signal processing. Exploiting an human visual perception model, we developed a perceived contrast measure in the DCT domain. Based on the measure, we designed a filter to make images seem to be vivid as well as finely resolved without boosting up noise. And after investigating how blocking and ringing artifacts occur during processing image signals, we devised a scaling filter that avoids such artifacts. The objective and subjective evaluations verify that the proposed method consistently produces competitive performances.