No-reference Blur Image Assessment Based on Local Sharpness

In order to effectively assess the quality of the image, the paper proposes a novel assessment method without the reference image that effectively combines the dual tree complex wavelet transform with the local sharpness of the image on the basis of the local sharpness of the image. Firstly, the image is decomposed multilayers by the dual tree complex wavelet and 6 different directions of the image are reconstructed, then the local sharpness of each direction is computed. Finally, the final local sharpness is computed by the linearly fusion with the weight. In 4 publicly available image databases, experimental results show that the proposed method is superior to other 6 kinds of the state-of-the-art methods in the blur images from the above databases and testify that the proposed method is closer to the human vision.