期刊
COMPUTATIONAL VISUAL MEDIA
卷 7, 期 4, 页码 453-466出版社
SPRINGERNATURE
DOI: 10.1007/s41095-021-0216-x
关键词
image retargeting; reconstruction; deep seam carving; image resizing
This paper introduces a novel method of image resizing in feature space using deep neural network layers. By adjusting image feature maps directly, important semantic information can be preserved while reducing noticeable artifacts.
Traditional image resizing methods usually work in pixel space and use various saliency measures. The challenge is to adjust the image shape while trying to preserve important content. In this paper we perform image resizing in feature space using the deep layers of a neural network containing rich important semantic information. We directly adjust the image feature maps, extracted from a pre-trained classification network, and reconstruct the resized image using neural-network based optimization. This novel approach leverages the hierarchical encoding of the network, and in particular, the high-level discriminative power of its deeper layers, that can recognize semantic regions and objects, thereby allowing maintenance of their aspect ratios. Our use of reconstruction from deep features results in less noticeable artifacts than use of imagespace resizing operators. We evaluate our method on benchmarks, compare it to alternative approaches, and demonstrate its strengths on challenging images.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据