一种改进的Seam Carving图像重定向算法

An Improved Seam Carving Image Retargeting Method

内容敏感图像重定向的核心思想是根据图像中不同内容的重要程度进行不同的缩放操作,目的是保留图像中重要的内容和结构的同时最大程度的减少视觉失真,Seam Carving是目前主流方法之一.传统的Seam Carving算法通过梯度图生成重要度图,只能检测出图像的边缘信息.本文提出基于梯度图和视觉显著度图共同生成重要度图的方法,既增强了重要区域的边缘信息,又很好的突出了视觉主体,同时又保护了重要区域的完整性.实验证明本文提出的方法在减少缩放图像视觉失真的同时,较好的保留了原图中重要区域的内容与结构.

The main idea of content-aware image resizing is performing different operations according to the importance of each pixel,and the goal is to preserve the important areas and reduce visual distortion.Seam carving is one of the main methods of image resizing. Traditional seam carving generates the importance map by gradient, which can only indicate high energy near the edges of an object.We propose a novel method combining gradient map and saliency map to generate an importance map,which can protect the edge information and visual prominent object,and preserve the integrity of the important areas. The experimental results demonstrate that the proposed method can reduce the visual distortion and preserve the content and structure of the original image.