Midpoint line algorithm for high-speed high-accuracy-rotation of images

Image rotation is a basic operation in most digital image processing systems as well as a necessary component in some visual pattern recognition systems. The speed and/or accuracy of image rotation are significant in most cases. In this paper, a new one-pass method, called the midpoint line algorithm, for realizing high-speed high-accuracy rotation of images is proposed. The new method, based on a novel view of images and their rotation, employs the principle of the midpoint technique for line generation. It mainly uses integer addition, increment by 1, and logic justification, and it significantly reduces the floating-point computations, so that it can be performed very fast without loss of high accuracy. Analysis of computation and accuracy of the method is given in comparison with those of the multipass methods.