Color Video JND Model Using Compound Spatial Masking and Structure-Based Temporal Masking

Abstract

In order to provide high transmission efficiency and visual quality for the color image/video in internet, effective estimation of its inherent just noticeable distortion (JND) is continuously an important topic. In this paper, a discrete cosine transform (DCT)-based perceptual model for estimating spatio-temporal JNDs of color videos is presented. Firstly, the compound masking adjustment integrated with different spatial masking factors is proposed. Based on the base detection threshold for DCT coefficients in luminance and chrominance components of the color image, the adjustment combined by luminance masking, pattern-based contrast masking, and cross masking is exploited to measure visibility thresholds for luminance component, while the adjustment induced by the variance-based statistical properties is utilized to measure visibility thresholds for chrominance components. Then, the local temporal statistics of luminance component are considered to design the structure-based temporal masking adjustment for further estimating visibility thresholds of video signals. To verify the proposed method, a subjective viewing test is designed and a fair rating process of evaluating the visual quality is carried out under the specified viewing condition. Experimental results show that the proposed method is able to measure more visual distortion tolerance than the existing color-based methods at high visual quality.

Keywords

• Just noticeable distortion (JND)
• spatial masking
• temporal masking