A NEW DCT-BASED JND MODEL OF MONOCHROME IMAGES FOR CONTRAST MASKING EFFECTS WITH TEXTURE COMPLEXITY AND FREQUENCY
Sung‐Ho Bae, Munchurl Kim · International Conference on Image Processing · 2013
Just noticeable distortion (JND) refers to minimum distortion threshold, which is visible to human visual system. JND models are applicable to various image and video processing fields, especially perceptual video compression fields by eliminating perceptually redundant amounts of signal components. The conventional DCT-based JND models for contrast masking (CM) effects simply reflect the texture complexity of images into a limited number of texture complexity categories. So such JND models are discrete functions of texture complexity, which are not sufficiently accurate. In this paper, the JND for the CM effects (CM-JND) is elaborately modeled, based on psychophysical experiments, as a 2D continuous function of DCT frequency and texture complexity that is measured by edge pixel density. The subjective performance evaluation shows the proposed CM-JND model makes the JND profile more consistent with real visual perception for the test images than the one with the conventional CM-JND models.