The Image Sharpness Metric via Gaussian Mixture Modeling of the Quaternion Wavelet Transform Phase Coefficients with Applications

In this paper, we make use of Gaussian mixture model (GMM) to describe the coefficients distribution of the quaternion wavelet transform (QWT). Derived from the parameters in GMM, the metric is proposed to find the relationship between the image blur degree and the distribution histograms of high frequencies coefficients. Also, the metric can be applied to smooth patch detection. Finally, experiments are conducted on natural images and the reasonable results indicate that the proposed metric can exhibit better performance than three common global sharpness measurements and satisfy the visual perception in the local smooth patch detection.