No-reference picture quality evaluation model using blockiness reduction algorithm

摘要

Most existing metrics require a reference together with the processed image or video in order to evaluate the visibility of these artifacts. This imposes obvious limitations on the applications that such a metric can be used for. No-reference metrics are much more powerful. No-reference signifies that the metric is not relative to the original but is an absolute value associated to a given image or video sequence. The image restoration algorithm such as blockiness reduction can be used to recover the original image and therefore a full-reference metric becomes applicable. In this paper, to develop a no-reference perceptual quality evaluation model, we adopt the combination with image restoration algorithm and full-reference evaluation model. If the restored image is improved to the same level as the original image, the full-reference model demonstrates high performance very much. Therefore, the performance of the proposed no-reference model has the possibility to be decided by the collaboration of the image restoration algorithm and the full-reference model.