Self-Supervised Denoise Method
Self-Supervised Denoise Method:Introduction使用noise 和cln作为GT的图像训练对比较难获得，延长曝光时间多次拍摄得到的AVE cln img也不一定是很好的cln GT，这种拍摄困难，要求拍摄主体不移动(人像眨眼睛，颤抖等)，在某些情况下在实验环境可
...
Coordinate Attention Paper Reading
Posted on
Coordinate Attention for Efficient Mobile Network DesignCVPR2021
Motivation:Baseline: SE attention, CBAMCBAM： 就是CA以及SA的结合attention。
Pipeline对比：前面的两者都没
...
VIT Paper Reading
Posted on
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale将transformer引入到了Image，代替了Conv，只是用transformer.之前的难点：
相较于words token, 图像的每一个p
...
CycleISP Paper Reading
Posted on
CycleISP: Real Image Restoration via Improved Data SynthesisMotivation:
CNN在AWGN 的数据集上表现很好，但是在real noise 中表现差，其原因是没有考虑到ISP将noise转变了，不在符合假设的高斯噪声。
Noise
...
MIRNet Paper Reading
Posted on
Learning Enriched Features for Real Image Restoration and Enhancement
The MRB whose main branch is dedicated to maintaining spatially-precise high-res
...
Local Relation Networks for Image Recognition Paper Reading
Posted on
Local Relation Networks for Image Recognitioncode左图为LR, 右图为Non-Local.
Geometry Prior可学习的先验特征。以某点为中心 $k \times k$ 大小的区域的到的几何先验，通过两个 $1\times1$ 卷积 中间
...
MM21
Posted on
总结下MM21投稿的工作(Multi-Decoder-based Patch-Adaptive Denoising Network)
一些小bug：framework中PFM名称没有修改
sup中表有一个参考文献为 ？
分析两个问题：
MDPA-Net 在SIDD上面能达到39.83,但是
...
MPRNet Paper Reading
Posted on
Multi-Stage Progressive Image RestorationMulti-stage networks are effective in high-level vision problems such as pose-estimation, action recognizatio
...
Deep Stacked Hierarchical Multi-patch Network for Image Deblurring
Posted on
Deep Stacked Hierarchical Multi-patch Network for Image Deblurring与之前方法pipeline的对比：(a) (b) 缺点：Increasing the network depth for very low-resolution inp
...
Pseudo-ISP Paper Reading
Posted on
Pseudo-ISP: Learning Pseudo In-camera Signal Processing Pipeline from A Color Image Denoiser这文章写的很混乱，读着不容易懂， 但是收获很多。
图像的噪声在raw域可以看作与空域无关，将图像投影到raw域，在r
...