Abstract: In the process of image resizing, seam carving algorithm causes damage to the geometry characteristics of images. To overcoming this shortcoming, an improved seam carving algorithm based on edge detection is proposed. First, seam is calculated and image contour feature is detected, and then, points of intersection are got. At last, energy around the intersection is upgraded to disperse over-concentrated seams. Simulation results demonstrate that the proposed approach can effectively reduce the image structure distortions caused by seam carving algorithm and well preserve the image prominent geometry features.

Key words: image resizing, geometric characteristics, edge detection, low-energy line, aberration