A Perceptually Optimized Wavelet Foveation Based Embedded Image Coder and Quality Assessor Based Both on Human Visual System Tools
Abderrahim Bajit, Mohammed Nahid, Mohammed Benbrahim, Ahmed Tamtaoui · 2019
In this paper, we propose a visually optimized embedded foveation image coder thereafter called VOEFIC. It introduces a visual weighting model to wavelet coefficients prior to control SPIHT encoding algorithm. It aims reach a targeted bit rate with a perceptual quality improvement with respect to a given bit rate, observation distance and a fixation point that determines the region of interest ROI. Our VOEFIC coder incorporates a battery of vision models including various masking effects related to the human visual system (HVS) perception. Thus, the coder deploys the visual model to weight the original wavelet coefficients, reshape its frequency spectrum, adapts its contents, eliminates its redundancy and finally increases the perceptual quality with respect to given coding parameters. The visual weighting model within wavelet sub-bands uses successively the following visual models 1) foveation masking applied to remove or at least to reduce high frequencies around peripheral regions 2) Luminance and Contrast masking deployed to adapt lightening and correct the contrast with to the wavelet just noticeable distortion thresholds. The paper also introduces a new objective quality metric deploying a visual model that integrates the human visual properties. The visual quality assessor offers a score based on a foveation probability scale FPS evaluate objectively the coding quality. The new perceptually optimized codec has the same complexity as the standard SPIHT coder. However, its visual results show that VOEFIC coder exhibits very good visual performances in terms of quality assessment index.