Blind image quality assessment based on fractal description of natural scenes

Motivated by the observation that visual perception is quite sensitive to the irregularities of natural scenes, the incorporation of the multi-fractal spectrum and the fractal dimension into blind image quality assessment for perceptual features extraction is introduced. On the basis of a box-counting method, the multi-fractal spectrum and fractal dimension are extracted from natural scenes and then their discrepancies are quantified. Experiments on the LIVE image database show that consistency with subjective evaluation is achieved. In addition, these are remarkable advantages when compared with other popular methods are demonstrated.