Steering核回归是一种自适应的、有效的图像恢复方法,在图像去噪、放大和去模糊中都得到了广泛应用。但此模型以高斯函数为核函数,故得到的恢复图像边缘,尤其是细小边缘常常会因过分平滑而模糊。提出基于鲁棒统计的各向异性核回归图像恢复模型,该模型在Steering核回归模型基础上,结合各向异性距离,以鲁棒统计权函数代替高斯核函数。大量图像恢复实验结果显示,与Steering核回归方法相比较,所提出方法得到的恢复图像质量显著提高,尤其是在细小边缘保持方面更具有明显优势。

Steering kernel regression is an adaptive and effective image restoration algorithm,which has been widely used for image denoising,enlargement and deblurring.However this model is based on the Gaussian kernel function,which often blurs the image edges,especially micro-edges in its restoration.A new anisotropic image restoration model based on robust statistics is proposed,which improved the Steering kernel regression model.The new method is to incorporate anisotropic distance and introduce robust estimation kernel function instead of Gaussian function.Extensive experiment results demonstrate that the new method can yield superior performance to that of the steering kernel regression,especially in preserving the details of the image edges.