一种基于边缘预测的图像实时放大技术

A real-time image magnificationtechnology based on edge prediction

为解决图像放大过程中的边缘模糊和边缘锯齿问题,提出了一种基于边缘预测的图像放大算法。该算法把图像划分成多个子区域单元插值,根据原始图像子区域单元像素相关性来预测放大图像的边缘,从而预测出新图像的非边缘像素点和边缘像素点。在此基础上非边缘像素点采取分区域线性插值,边缘像素点采取对角线插值等方法,这样能有效地抑制边缘模糊和边缘锯齿现象。该算法可以实现对灰度图像的在一定范围内的任意放大并且能够得到比较理想的结果。通过对比其他算法,以及在硬件系统上实现后的结果分析,表明该算法明显优于其他几种图像放大算法,并且计算复杂度较低,仅仅比双线性插值稍微复杂,可用于视频图像的软硬件实现。

To solve the problem of edge blurring and edge algorithm based on edge prediction was proposed. In the jaggingIn image zooming process,a new image magnification algorithm, an image was firstly divided into a few sub - region units and then edges of the magnified image were predicted according to the correlation of the pixels in sub - region units of the original image, from which both the non - edge pixels and the edge pixels of the desired image could be found. Finally, the non - edge pixels and the edge pixels were calculated by liner interpolation in different regions and diagonal interpolation respectively, which effectively suppress the edge blurring and edge jagging. The algorithm could magnify a gray image to any times within limits with acceptable result. After comparing with some other algorthms and testing the practical performance of the algorithm in hardware device it turns out that the algorithm has clear advantages over the other algorithms in terms of quality and the relatively low time complexity which is slightly higher than the time complexity of the bilinear. Consequently, the algorithm can be used for both hardware and software solutions to video image.