Single-Image Super-Resolution via Linear Mapping of Interpolated Self-Examples笔记
2016-08-16 14:51
483 查看
此文2014年TIP。
主要points
double pyramid: a traditional single pyramid and a side pyramid of interpolated levels(figure 5)
对traditional single pyramid,是对input LR image I_L作为I_0多层下采样 I_-n。
对 a side pyramid of interpolated levels, 是对上述traditional pyramid 进行插值(scale = p)
则在Fig.5中,左侧pyramid视为高分辨率image,右侧视为下采样再插值得到的same size 的 image.
利用DM(direct mapping)就可以得到映射关系,将低分辨x_i^l恢复为x_i^h。
但是,scale 为 p^n = 1.25^n, 通过层层渐进的方式得到想要的放大倍数。这里面还有一点就是,每次重建之后就把当前重建的样本利用IBP约束,然后还要将结果加入double pyramid作为下次scale渐进的dictionary.
drawback: 文章中说要通过gradual upscalings 来SR,而NP = log_p(s)可能不是想要的放大倍数(不是整数),需要对结果调整大小,但是文章没说具体如何调整。
第一次写,试下效果==
主要points
double pyramid: a traditional single pyramid and a side pyramid of interpolated levels(figure 5)
对traditional single pyramid,是对input LR image I_L作为I_0多层下采样 I_-n。
对 a side pyramid of interpolated levels, 是对上述traditional pyramid 进行插值(scale = p)
则在Fig.5中,左侧pyramid视为高分辨率image,右侧视为下采样再插值得到的same size 的 image.
利用DM(direct mapping)就可以得到映射关系,将低分辨x_i^l恢复为x_i^h。
但是,scale 为 p^n = 1.25^n, 通过层层渐进的方式得到想要的放大倍数。这里面还有一点就是,每次重建之后就把当前重建的样本利用IBP约束,然后还要将结果加入double pyramid作为下次scale渐进的dictionary.
drawback: 文章中说要通过gradual upscalings 来SR,而NP = log_p(s)可能不是想要的放大倍数(不是整数),需要对结果调整大小,但是文章没说具体如何调整。
第一次写,试下效果==
相关文章推荐
- 【超分辨率】Enhanced Deep Residual Networks for Single Image Super-Resolution
- 论文阅读之:Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
- 每日论文Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
- “Single Image Super-resolution using Deformable Patches”
- Enhanced Deep Residual Networks for Single Image Super-Resolution
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
- Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural
- Hyperspectral image super-resolution via non-local sparse tensor factorization
- Enhanced Deep Residual Networks for Single Image Super-Resolution
- 《Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network》阅读笔记
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
- Hyperspectral Image Super-Resolution via Non-Negative Structured Sparse Representation论文阅读笔记(PART)
- Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
- 《Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network》阅读笔记
- 《Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network》阅读笔记
- 论文阅读笔记:Image Super-Resolution Using Deep Convolutional Networks
- PAISR:Rapid and Accurate Image Super Resolution 详解(内附代码code)
- Rails mapping array of hashes onto single hash
- Learning a Deep Convolutional Network for Image Super-Resolution—Chao Dong_ECCV2014
- 人群计数--Single-Image Crowd Counting via Multi-Column Convolutional Neural Network