Edge Curve Scaling and Smoothing with Cubic Spline Interpolation for Image Up-Scaling

Image up-scaling is an important technique to increase the resolution of an image. Earlier interpolation based approaches have low computation complexity while cause blurring and ringing artifacts in edge regions due to the loss of high frequency details. Patch-based super resolution achieves satisfactory up-scaling images at the penalty of high computation cost. In this paper, we present a scalable edge map to recover high frequency components of edge regions in up-scaled images to improve the sharpness and use a range compression method to reduce ringing artifacts. We propose an efficient 1-D and 2-D approaches to extract edge curves from an original image. Then the cubic spline interpolation is adopted to up-scale an edge map. A smooth function is added to remove zigzag, stair, trapezoid artifacts. Then we apply the patch-based method only on up-scaled edge map to recover the high frequency components. The execution time of our proposed method is only 10 % to 5 % compared to the multi-resolution patch-based super resolution method. Experimental results show that we can achieve similar image quality with G. Freedman et al.’s method [ 19 ].