图像局部纹理特征自适应超分辨率重建

An Adaptive Super-resolution Reconstruction of Image Local Texture Feature

在双边总变分(BTV)正则化方法中,由于同时考虑了周围像素与中心像素的几何距离和灰度相似性,获得了比Tikhonov正则化方法和总变分(TV)正则化方法更好的重建质量。然而,在BTV方法中,由于正则化参数λ为一个定值,使得该方法不能同时保持图像的边缘纹理信息和抑制图像噪声。针对这个问题,提出一种图像局部纹理特征自适应的正则化重建方法,基于灰度共生矩阵提取图像局部纹理特征,建立正则化参数与图像局部纹理特征的函数关系,使正则化参数λ随图像局部纹理特征自适应调整。实验结果显示,与BTV方法相比,该方法能使图像的边缘和纹理细节重建效果更好,并有效抑制噪声。

In bilateral total variation regularization method（BTV）, considering geometric distance and gray level similarity of the center pixel and the surrounding pixels, the method to get better reconstruction quality than Tikhonov regularization method and total variation regularization method （TV） is obtained. However, in BTV meth？od, the regularization parameterλis a fixed value, so the method cannot maintain the image edge texture informa？tion and suppress image noise at the same time. In order to solve this problem, an adaptive regularization reconstruc？tion algorithm for image local texture feature is proposed, and based on gray level co-occurrence matrix（GLCM）, the image local texture feature is extracted, the function relationship of regularization parameters and image local texture feature is established, so regularization parameterλis adjusted adaptively according to image local texture feature. The experimental results show that compared with BTV, this algorithm can better reconstruct the image edge texture details and suppress the noise effectively.