In this paper, a reduced-reference image quality assessment metric is proposed, which measures the difference of the regularity of the projection spatial arrangements between the reference image and the distorted image. The projections of an image are first extracted by Radon transform. Then fractal dimensions are computed on each projection and concatenated as the image features that characterize the image structures from the view of the spatial distribution. Finally the image features are pooled as the quality score using ℓ1 distance. The proposed approach was evaluated on four public benchmark databases. Experimental results have demonstrated the excellent performance of the proposed approach.