An image retargeting scheme with content-based cropping and local significance aware seam carving

This research presents an image retargeting scheme consisting of two main operations, the content-based cropping and seam carving, to modify the width/height of imagery data. The seam carving is first employed to remove insignificant pixels in the middle part of image. The selection of seams is based on the gradient in one direction and local energy to avoid obvious distortions. When the background tends to be uniform, some seams can be inserted in a similar way to make the aspect ratio closer to the target one. The content-based cropping that considers the visual saliency is then applied to remove boundaries. The final image is formed by scaling directly if necessary. The experimental results will demonstrate the advantages of the proposed method.