Edge-directed interpolation based on Canny detector

This paper presents a modified method of the new edge-directed image interpolation algorithm. The basic idea is to first estimate the properties of the pixels to be interpolated and then calculate the unknown values based on different interpolation formulas. The improvement is that we propose a sample-points-selection from a statistical standpoint before interpolation to solve the problem of covariance mismatch in high frequency regions, and a special interpolation scheme on edge pixels based on Canny detector to avoid edge blurring. Experimental results show that our algorithm performs higher objective quality when applied to high-contrast image.