基于粗糙集约简的图像插值方法

Image interpolation scheme based on rough sets reduction

提出了一种基于粗糙集约简的支持向量机图像插值方法,目的在于提高基于学习的插值方法的插值效率,改善放大图像边缘模糊现象。首先在原始图像上利用已知的像素灰度值及邻域内像素间的相关性构造训练样本集;然后利用粗糙集约简算法约简掉其中重要度较小的特征,并用约简后的样本集训练支持向量机;再用测试样本及训练好的支持向量机估计偶行偶列的像素灰度值;最后利用测试样本及训练好的支持向量机估计剩余的未知像素灰度值。仿真表明,所提方法有效提高了插值效率,获得了较好的客观指标,得到了满意的插值图像。

In order to obtain visually pleasing image, this paper proposed an image interpolation method. It constructed training sample set based on the original image using the known pixel gray values and the correlation within the neighborhood pixels. Next, it used rough sets to reduce the training sample set, and trained the support vector machine with the reduced training sample set. Then, it estimated the pixel gray values in even row even column using the trained support vector machine. Finally, it estimated the pixel gray values in odd row even column and in even row odd column using the trained support vector machine. Simulation results show that the proposed method increases the efficiency of interpolation methods, improves the magnified image edge blurring, and obtains better objective indicators.