中文摘要：

传统的插值算法通常会使目标图像边缘模糊,而基于2×2或4×4邻域的局部自适应算法虽然能保持图像边缘特征,但算法一般过于复杂,不便于硬件实现。提出的改进算法不仅对LAZA算法流程进行优化,而且使用中值滤波代替分类加权平均算法,既提高算法效率,也能保持图像的边缘特征。实验结果表明该算法在放大采用双三次内插获取的1/2图像有更好的效果,并利用边缘图像相关系数进一步验证该算法在保持图像的边缘特征方面优于传统插值算法。

英文摘要：

Traditional interpolation algorithms usually will blur the targeted image edges,and though the locally adaptive algorithm based on 2 × 2 or 4 × 4 neighbourhood can maintain the image edge feature,but the algorithm is generally too complex,it is not easy for hardware implementation. The improved algorithm proposed in this paper not only optimise the flowchart of LAZA algorithm,it also uses the median filtering algorithm to replace the classified weighted average algorithm,which improve the efficiency of the algorithm,and also preserves the edge features of image. Experimental results show that this algorithm has better effect in amplifying the 1 /2 image obtained using bicubic internal interpolation. Moreover,we further verify by using the correlation coefficient of edge image that the proposed algorithm outperforms the traditional interpolation algorithm in preserving the image edges.