传统的图像插值算法存在边缘模糊和边缘锯齿，已有改进算法改善了插值图像质量，但存在斜边缘锯齿或局部扭曲变形等问题。为解决上述问题，提出了改进的图像自适应梯度插值。该方法首先根据图像的局部不对称性和局部梯度特征在1维方向上修正插值点空间距离，并将1维修正结果合并到2维空间，然后将修正的空间距离应用到传统的插值算法中。实验结果表明，该方法改善了插值图像的峰值信噪比，有效保护了图像在各个方向上的边缘信息。

Conventional image interpolation introduces blur and jagged edges. Existing improved interpolation algorithm obtains better visual quality of interpolated images than traditional interpolation algorithm. But interpolated images may still introduce jagged edges or distortion around the edge. To solve the problem, the paper proposes an improved adaptive image interpolation algorithm with gradient features. The basic idea is to first correct the special distance of interpolated pixels along one dimension on the basis of the local asymmetry features and gradient features of the original images, convert the corrected results of one dimension into two dimensions and then apply the corrected distance to conventional interpolation algorithm. Experimental results demonstrate that the improved method produces good results in regard to the peak signal-to-noise ratio(PSNR)and succeeds in preserving interpolation image edges of various directions.