Reduced reference Image Quality Assessment for transmitted images using digital watermarking

摘要

Objective Image Quality Assessment (IQA) aims to automatically measure the quality degradation perceived by human eyes. In this paper, a new reduced-reference image quality assessment algorithm for channel distortion (distortion specific metric) has been proposed. This algorithm uses a small part of the original image (as a reduced-reference) and embeds this part, in the image, as a watermark. At the receiver side, this part is extracted from the watermarked distorted image and with the help of received part of the image and the extracted watermark the channel is estimated. In this method, neither receiver requires to know the training sequence in advance nor the sender requires to send the training sequence, and channel can still be estimated using digital watermarking. Once the channel coefficients are estimated, the condition number of the channel convolution matrix gives a direct measure for quality of a received image. Proposed algorithm can simultaneously be used for usual watermarking applications and image quality assessment.