基于全局图像结构信息的Seam Carving算法*

本文提出了一种基于全局图像结构信息的Seam Carving算法，它根据像素的重要性修改图像尺寸和比例。通过从图像提取特定方向的边缘结构信息，再利用每个像素的梯度信息，从全局和局部两方面定义新的像素能量计算函数，以此来阻止Seam通路与特定方向图像边缘的交叉，避免边缘像素的不一致位移，以此保持图像的边缘结构。实验结果证明，本文算法减少了处理后图像的结构形变，有效改进了Seam Carving算法的处理效果。

Seam Carving algorithm based on global image structure

In this paper, we present a seam carving algorithm based on global image structure, which changes the sizes and scales of images by the importance of each pixel. We define a new pixel energy function from both global and local information by extracting special image edges and computing each pixel gradients, which prevents seams from crossing image edges, avoids the pixels on edges moving differently and preserves image edge structures. Experiments confirm that our algorithm reduces the structure distortion and improves the visual effect of images by seam carving process effectively.