摘要： 归纳图像调整过程中的相似性判据，提出一种基于相似性判据的图像调整算法。采用Seam Carving算法按一定调整量分步进行 图像尺寸调整，利用相似性判据判断调整图像的变形程度，当变形达到一定程度时，换用双向相似度迭代优化算法完成图像剩余尺寸的调整。实验结果表明，在图像调整量很大时，利用该图像调整算法仍可较好地维持图像的整体视觉效果。

关键词: 图像尺寸调整, 相似性判据, 非相似度, 双向相似度迭代

Abstract: The similarity criterion for image resizing is concluded, and a image resizing algorithm based on similarity criterion is proposed. Image resizing is implemented by Seam Carving algorithm, it removes the seams with the lowest gradient energy. The degree of image deformation and similarity can be judged using the similarity criterion. Before the degree of deformation reaches a threshold, the resizing method is changed to bidirectional similarity iteration optimization for residual amount. Experimental results show that, even though the amount of adjustment is large, the algorithm can preserve global visual effect.

Key words: image resizing, similarity criterion, dissimilarity, bidirectional similarity iteration

中图分类号: 

• TP391