A Novel Blind JPEG Image Quality Assessment Based on Blockiness and the Low Frequency Feature in DCT Domain

Blind image quality assessment metrics play an important role in the field of image processing. Blind image quality assessment methods, which are specific to a given type of distortion, are very popular for different image processing applications. JPEG compression is one of the most common image compression methods. In this paper, a support vector regression approach is adopted to assess the quality of JPEG compressed images without reference image. At first, the low frequency feature in DCT domain and the blockiness feature are calculated to present the distortion information of an image. Second, the JPEG dataset is divided into training and testing set randomly. The training set is used to build the SVR model, and the testing set is used to predict the quality score. Finally, combining with MOS or DMOS, the quality score is predicted by SVR model. Extensive experiments on LIVE database demonstrate that the proposed method outperforms the state-of-art methods both on predict accuracy and computational complexity.