Improved seam carving for image retargeting with sift feature preservation

Abstract

This paper presents an effective and simple image resizing method. Our method is an improved version of seam carving that changes the backtracking basis of seam carving. We use a Scale Invariant Feature Transform (SIFT) feature in our method. SIFT key points are mainly located on high-contrast regions of an image. By using saliency considering SIFT as our backtracking basis, a resized image can preserve the SIFT features of the original image. Therefore, the resized image can be more visually acceptable than the image resized by traditional seam carving.