HVS based iterative hard thresholding recovery for compressive sensed image

摘要

Iterative hard thresholding (IHT) is a technique popularly used in compressive sensing recovery. Human eyes have different sensitivity depending on frequency, thus, errors in low frequency affects human perception more seriously than in high frequency. However, IHT treats all transformed coefficients equally without differentiating their visual sensitivity. This paper proposes an improved IHT scheme based on human visual sensitivity by using a weighting process. Visually significant coefficients are emphasized more by large weights before the hard thresholding. Experimental results show its superiority over conventional IHT in both objective and subjective qualities.