In ground-based large aperture solar telescopes, the speckle image reconstruction technique combined with adaptive optics (AO) correction is generally used to get near diffraction-limited solar images. In order to select the high quality images from the AO-corrected high resolution solar image sequence, an automated no-reference image quality assessment (IQA) is needed. According to the noise characteristics of solar AO images, an IQA metric based on an image power spectrum and human visual system is developed. By the incorporation of noise masking and shifting the spatial frequency range of summation, our IQA metric could select sharper images with less noise than previous works based on the image power spectrum, even if there is image scaling. Compared with existing general-purpose IQA metrics and previous metrics specifically designed for solar IQA, experimental results verify that the proposed metric gains better performance whether on robustness to blur and noise, or on selecting high quality frames containing solar granulations, sunspots, or both of them.