The Correlation Dimension: A Video Quality Measure

Correlation dimension is a measure of the multidimensional complexity of an object. Stemming from the area of chaos theory and having several applications involving the study of the convergence and the recurring patterns of random signals, it has been proven to be a possible way to assess video quality. Based on its meaning in the multidimensional space of color fractals, it can be used, in the context of a fractal’s intrinsic similarity to natural shapes and colours, to quantify the aesthetic and harmonic properties of an image. Our approach in the assessment of the perceived quality of a video stream is based on the analysis of the fractal dimension of video signals expressed in the CIE L*a*b* color space. This colour space has a strong resemblance to the human visual perception system, thus making its Δ

E

2000

norm relevant for the measurement of the perceptual difference between colours, and hence useful for image quality assessment. The fractal dimension is computed through the correlation dimension definition. In this paper we expose the experimental results obtained in a simulation of a real-life scenario: the streaming of a video of a football game over a busy network.