| With rapid development of information technology, digital multimedia technology has booming too. In many techniques of digital image processing, image rotation is one of basic image processing operation that is widely used in many fields such as medical imaging, radiology and digital photography. All these applications require the preservation of the integrity of the image information. Ideally,if an image is rotated by a certain angle and then it is rotated by the opposite angle, the resulting image should be exactly the same as the original image. However, all proposed image rotation methods so far incur information loss and thus cannot meet such requirement. This problem of information loss in general should be minimized in diverse applications. For example, in medical imaging, the loss of information during image rotation may cause the loss of small but vital features that cause misdiagnosis. So how to better preserve the information in original image after image rotation operation become one problem that many researches study.Previous image rotation algorithms are mostly based on the interpolation theory, that is to say, the accuracy of interpolation methods decide the quality of image rotation algorithms. In the traditional interpolation algorithms, nearest neighbor and bilinear interpolation approaches are extremely simple to implement but they have the tendency to produce undesirable artifacts. Bicubic interpolation can preserve finer details than above two methods and it becomes a standard application of some commercial image editing programs such as Adobe Photoshop and Corel Photopaint. But this method has higher computational complexity so that it become impractical for higher-order models. In addition, some researchers proposed one rotation method that based on the shear principle, this method has three steps to implement image rotation with hardware. One recently proposed image rotation algorithm using hermite expansions has more better accuracy than above three methods, but the stability of this approach is not ideal.In this work a radial basis function(RBF) interpolation method is applied to digital image rotation because this approach implements simple and gets high accuracy in approximation problem. There are two advantages of this approach.One is that by using RBF 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. |
