人工智能培训

搜索

深度学习论文:学习视差注意立体图像超分辨率(Learning Parallax Attention for Stereo Image Super-Resolut

[复制链接]
hhtonyhh 发表于 2019-3-15 12:00:16 | 显示全部楼层 |阅读模式
hhtonyhh 2019-3-15 12:00:16 802 0 显示全部楼层
深度学习论文:学习视差注意立体图像超分辨率(Learning Parallax Attention for Stereo Image Super-Resolution)立体图像对可用于改善超分辨率(SR)的性能,因为从第二视点提供了附加信息。然而,由于立体图像之间的差异显着变化,因此将该信息合并到SR是具有挑战性的。在本文中,我们提出了一种视差注意立体超分辨率网络(PASSRnet)来整合来自SR的立体图像对的信息。具体而言,我们引入具有沿着极线的全局感受场的视差注意机制来处理具有大的视差变化的不同立体图像。我们还提出了一个新的和最大的立体图像SR数据集(即Flickr1024)。广泛的实验表明,视差注意机制可以捕获立体图像之间的对应关系,以提高SR性能,同时计算和内存成本最低。比较结果显示,我们的PASSRnetarts在Middlebury,KITTI 2012和KITTI 2015数据集上具有最先进的性能。
Stereo image pairs can be used to improve the performance of super-resolution(SR) since additional information is provided from a second viewpoint.However,it is challenging to incorporate this information for SR since disparitiesbetween stereo images vary significantly.In this paper, we propose aparallax-attention stereo superresolution network (PASSRnet) to integrate theinformation from a stereo image pair for SR.Specifically, we introduce aparallax-attention mechanism with a global receptive field along the epipolarline to handle different stereo images with large disparity variations.We alsopropose a new and the largest dataset for stereo image SR (namely, Flickr1024).Extensive experiments demonstrate that the parallax-attention mechanism cancapture correspondence between stereo images to improve SR performance with asmall computational and memory cost.Comparative results show that our PASSRnetachieves the state-of-the-art performance on the Middlebury, KITTI 2012 andKITTI 2015 datasets.深度学习论文:学习视差注意立体图像超分辨率(Learning Parallax Attention for Stereo Image Super-Resolution) l1773wqj7s6s6Jfq.jpg
URL地址:https://arxiv.org/abs/1903.05784     ----pdf下载地址:https://arxiv.org/pdf/1903.05784    ----深度学习论文:学习视差注意立体图像超分辨率(Learning Parallax Attention for Stereo Image Super-Resolution)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

hhtonyhh当前离线
新手上路

查看:802 | 回复:0

快速回复 返回顶部 返回列表