A new image rotation approach using radial basis functions

Image rotation is a basic operation in many image processing and analysis tasks. It can be viewed as an image interpolation problem. In this paper, we propose a new approach to this problem using radial basis function (RBF). There are two advantages of this approach. One is that by using RBF basis functions locally, image local features can be very accurately captured, thus providing a highly adaptive interpolation mechanism. The other is due to the fact that RBF can be implemented simply as solving a traditional linear system with the computational complexity of O(N), where N is the total number of pixels. Experimental results show that the proposed method outperforms classical interpolation methods including the nearest, bilinear and bicubic interpolations and the more recently method using Hermite basis expansion.