基于局部多特征的低分辨率图像纹理增强技术研究

Research on low resolution image texture enhancement technology based on local multi⁃features

传统纹理图像增强方法存在轮廓模糊问题,导致增强效果较差。针对这些不足,提出基于局部多特征的低分辨率图像纹理增强技术。使用仿射变换方法查看近似平面单应性变化,结合RANSAC算法,匹配低分辨率图像纹理局部多特征点。提取低分辨率图像低频和高频信息,采用插值算法逐点划分低分辨率图像,得到像素点块。从中寻找匹配后最佳匹配块,采用高斯函数抑制重叠区域影像,将所有高频信息块拼接,得到低分辨率图像。将加窗及合成图像矩阵转换成同等大小隶属度矩阵,利用拉伸变化趋势强化轮廓,由此完成图像纹理增强。由实验结果可知,该技术在UMD数据集、UIUC数据集和Brodatz数据集下增强效果较好,能够得到准确完整的图像,有效恢复图像纹理细节。

The traditional texture image enhancement method has the problem of blurring contour,which leads to poor enhancement effect.To solve these problems,a texture enhancement technique based on local multi features is proposed.The affine transformation method is used to view the change of the approximate plane homography,and the RANSAC algorithm is used to match the local multi feature points of the texture in the low resolution image.Low and high frequency information of low⁃resolution image is extracted,and low⁃resolution image is divided point by point by interpolation algorithm to obtain pixel block.The best matching block is found,and the overlapped region image is suppressed by Gauss function.All high frequency information blocks are spliced to get low resolution image.The windowed and synthesized image matrix is transformed into membership matrix of the same size,and the contour is enhanced by stretching the changing trend,thus the image texture is enhanced.From the experimental results,we can see that the enhancement effect of this technology is better in UMD data set,UIUC data set and Brodatz data set,and it can get accurate and complete images,and effectively recover image texture details.