一种基于正则化的边缘定向插值算法

针对传统的基于线性回归模型插值算法不能对变化剧烈的边缘进行有效插值的问题,该文提出一种基于正则化的边缘定向插值算法。算法主要分为两部分:参数估计部分与数据估计部分。在参数估计部分,为了更加准确地描述图像局部结构,把已估计的高分辨率像素作为训练像素的一部分,用以进行回归模型参数的估计。在数据估计部分,引入像素平滑方向作为正则化项,以降低参数的误估计引起的数据估计偏差。实验结果表明,该算法能很好地保持图像的边缘特征,尤其在变化比较剧烈的边缘区域;与双三次插值算法及基于正则化的局部线性回归插值算法(Regularized Local Linear Regression,RLLR)相比,该算法能取得更好的视觉效果及较高的PSNR值。

An Edge Directed Interpolation Algorithm Based on Regularization

The traditional methods based on linear regression model preserve the edge in some degree, but hardly work on the sharp edge. To solve this problem, an edge directed interpolation algorithm based on regularization is proposed in this paper, which is composed of the parameters estimation part and the data estimation part. In the first part, the high resolution structures which have been estimated are taken as one part of the training pixel to estimate the parameters of the linear regression model for effectively describing the structure. In the second part, the smooth pixel’s direction is applied as the regularization to reduce the error of estimated data aroused from the incorrect parameters. Experimented results show that the proposed method preserves the edge of image effectively, and both the visual effects and the PSNR are all better than bi-cubic and Regularized Local Linear Regression (RLLR).