Perceptual Image Coding Based on Visibility Threshold Model in Wavelet Transform Domain

摘要

A visibility threshold model is established to describe the JND (just-noticeable distortion) quantitatively in wavelet domain, based on some of characteristics of HVS (human visual system). By using this model, we design a new visual quantizer. The quantized wavelet coefficients are grouped into 8x8 blocks and entropy coding in JPEG is used. Hence, the output bitstream of our coding scheme has the same structure as that in JPEG. Meanwhile, our image coding scheme eliminates the blockness artifacts in DCT.