摘要

A new algorithm for no-reference blurred image quality assessment is proposed. This metric is based on block-based discrete cosine transform statistics and linear prediction method. We compare the performance of the proposed method with four blur metrics on three famous databases and our work shows the best results.