【摘要】 提出了一种基于图像多方向最大相关性的数字图像插值算法.数字图像插值过程中不可避免会产生图像细节模糊和边缘锯齿化,同时在处理视频信号时要求有较低复杂度以满足实时信号处理.针对以上问题和要求,本文根据图像空域内邻近像素点多个方向的内容相关性和几何相似性,得到一种基于图像最大相关性的自适应优化插值算法.该算法能有效解决插值后图像的锯齿效应及边缘模糊等问题.实验结果表明,该方法插值后的图像信噪比性能明显优于双线性方法,插值倍率较小时接近两次立方法,随着插值倍率增大性能也优于两次立方法,且计算量不大,经适当改进后可以用于实时视频信号处理.更多还原

【Abstract】 An image interpolation algorithm based on max-relativity of multiple directions is presented in this paper.The smooth effect and staircase of image detail or edges are inevitable during image interpolation.In video signal processing the computational complexity of the algorithm is also important.This paper proposes an adaptive max-relativity optimum interpolation algorithm for digital image to resolve the problems above.The algorithm takes into account of the relativity and symmetry of local neighbors and thus removes the staircase and smooth effect of the interpolated image more effectively.Experimental results demonstrate that the PSNR of the image gained by our proposed algorithm outperforms bilinear,superior than bicubic when the zoom factor is larger and has less computational complexity.The algorithm is also suit for real time video signal processing by some correction.更多还原