基于小波的改进加权抛物线插值的图像超分辨率算法

Using improved weighted parabolic interpolation and wavelet transformation to zoom images for super-resolution

为了尽可能地保持图像的基本信息,提高图像的视觉效果和空间分辨率,提出一种基于小波的改进加权抛物线插值算法,即在传统的加权抛物线算法上增加插值的误差补偿项。利用sobel算子设定插值点的边缘方向,得到初始放大图像。利用小波变换提取高频成份,原始图像幅值增强充当低频部分,再经过小波逆变换得到高分辨率图像。实验结果表明,相对于传统的图像放大算法,该算法考虑到全局相关性,得到更加清晰的边缘信息。

In order to preserve the basic information of images and to improve their visual performance and space resolution, a novel image zooming algorithm is proposed by using improved weighted parabolic interpolation and wavelet transformation which incorporates the error-amended part into the classical weighted parabolic interpolation algorithm. Directions of interpolation points are determined by the Sobel operator, then a preliminary zoomed image is obtained. High-frequency components are got by a wavelet transformation, and low-frequency components are replaced by the enhanced amplitude of the original image. At last, a high-resolution image is achieved by inverse wavelet transformation. The experimental results indicate that, compared with traditional image zooming algorithms, this algorithm can get clearer and sharper edges due to considerations of global pixels correlation of the original image.