结合视觉显著图的Seam Carving图像缩放方法

Seam Carving image scaling method with visual significant graph

Seam Carving算法在缩放主体区域与背景颜色对比不明显或者视觉主体区域较大的图像时,可能会造成图像视觉主体变形和重要内容缺失的现象.为了改善Seam Carving算法的不足,采用图像梯度图和显著图结合的方法来改进图像的梯度能量图.实验表明,这种方法在缩放图像时比Seam Carving算法更能很好地保持图像重要内容,整体视觉效果较好,图像像素的平均能量值更大,图像缩放质量更好.

Seam Carving algorithm may result in the distortion of the visual subject area and the missing of the important content of the image when scaling an image with no obvious contrast between the subject area and the background color or with larger subject area.In order to improve the lack of Seam Carving algorithm,we use a combination of image gradient graph and saliency map to improve the gradient energy graph of the image.The results of experiments show that,compare with the Seam Carving algorithm,the proposed method can better maintain the important content of the image,preserves a better overall visual effect,and obtains a larger average energy value of the image pixel and a better image scaling quality in scaling the image.