Image sharpness assessment based on local phase coherence and LAD criterion

摘要

Edge preservation and precise restoration of details and sharp parts of images is an important issue in modern multimedia systems, 3D imaging, medical imaging and laser microscopy. Objective evaluation of image restoration algorithms in the sense of sharpness is crucial for choice of method to be used. Moreover, automated focus stacking, selective deblurring and other smart restoration methods rely on the proper sharpness measure. In this paper, a novel approach for image sharpness assessment is proposed. The approach is based on local phase coherence (LPC) in the complex wavelet domain. Using maximum pooling of eight LPC maps in eight spatial directions around each pixel, an image sharpness map is generated. Least absolute deviation criterion applied on sorted non-zero elements of the sharpness map results in the proposed global image sharpness measure. The proposed measure is better suited for applications, in comparison with competitive measures.