current position：Home>Computational Photography - Image Denoising (2)
Computational Photography - Image Denoising (2)
This chapter will be organized图像去噪相关的内容,主要包括：噪声类型,评估方法,传统去噪方法,Deep Learning Denoising Methods,常用数据集,This chapter mainly discusses the last two parts.The previous content can refer to：计算摄影——图像去噪（一）_Turned_MZ的博客-CSDN博客
Commonly used denoising datasets
At present, the main methods for establishing denoising datasets are as follows:3种：
- Obtain high-quality images from existing image datasets,然后做图像处理（线性变化、亮度调整等）And add artificially synthesized noise according to the noise model,生成噪声图像.However, this method is artificially synthesized due to noise.,Different from the noise in the real scene,Therefore, the effect is not good in the real scene.
- for the same image,Shoot low-sensitivity images as true values,High-sensitivity images as noise images,And adjust the exposure parameters to real two image brightness.This class of methods directly uses the low-sensitivity image as the ground truth,Inevitably there will be residual noise,And there may also be brightness differences and misalignment issues with noisy images.
- Shooting multiple images of the same scene in succession,Then a ground truth is synthesized by image registration, etc.,This way requires a lot of images to be taken,工作量比较大,But the final true value is of high quality.
Common datasets and corresponding papers：
|数据集||GT||number of scenes||Picture logarithm||主要领域||说明||相关论文|
拍摄了120dark scene,Contains indoor and outdoor scenes.About each scene4张图像,包含2One noisy image and two low-noise images.
|RENOIR - A Dataset for Real Low-Light Image Noise Reduction|
|Nam-CC15||均值||11||17||物品||包含11个场景,and mostly similar objects and textures.针对这11A total of scenes were shot500张JPEG图像.||A Holistic Approach to Cross-Channel Image Noise Modeling and its Application to Image Denoising|
|DND||低ISO||50||50||室内外||拍摄50个场景,Includes indoor and outdoor scenes.||Benchmarking Denoising Algorithms with Real Photographs|
|PolyU||均值||40||40||室内外||拍摄了40个场景,Including normal light and dark scene scene,Outdoor normal lighting scene.Continuous shooting for each scene500次||Real-world Noisy Image Denoising: A New Benchmark|
|SIDD||均值||200||400||多领域||用5个相机（Google Pixel、iPhone 7、Samsung Galaxy S6 Edge、Motorola Nexus 6、LG G4）Shot under four camera parameters10个场景,200scene instance,Each scene was shot continuously150张图像.其中160scene instances as a training set,40scene instances as a test set（the benchmark）.||A High-Quality Denoising Dataset for Smartphone Cameras|
Deep Learning Denoising Methods
Estimating Noise Residuals
in the field of information processing,Learning the amount of change in the signal is often simpler than learning the original signal,This idea is used in very efficient residual networks.DnCNNThe model draws on this idea,It does not directly output the denoised image,But the prediction residual image,The observation noise image and the difference between potential noiseless image.
CNN在VGG的基础上进行修改,网络结构是（卷积、BN、ReLU）级联的结构,Inside the model is not likeResNetThere is also a long-hop connection,Instead use residual learning at the output of the network.结构如下：
- 论文中强调了batch normalization的作用
noise estimation model
CBDNetThe model is a real image non-blind denoising model,对于RAW格式的图像,Its noise model is as follows：
image with real noise=图像信号L+real noise signaln(L).CBDNetModel using a noise estimation subnet estimate the noise level,Then together with the original input image input open-label denoising based on jump layer connection network,其结构如下：
1、The noise estimation sub-network converts the noise observation image to the estimated noise level image,Then enter it with the original image,Use the non-blind denoising sub-network to get the final denoising result,除此之外,Noise estimation operator network allows the user to estimate the noise level image input open-label alignment adjustment before denoising subnet,It proposes a simple strategy,即y = ay,i.e. linear scaling,This gives the model an interactive denoising capability.
盲去噪是指在去噪过程中,用于去噪的基础是从有噪声的样本本身学习来的.换句话说,无论我们构建什么样的深度学习体系结构,都应该学习图像中的噪声分布并去噪.所以和往常一样,这都取决于我们提供给深度学习模型的数据类型.The blind denoising, on the other hand.
2、Noise estimation network loss function：Asymmetric loss function（ asymmetric loss）+Total Variation Loss Function（ total variation）
2.1、Asymmetric loss function,Calculate the noise estimate map and noiseground truth的平方差：
其中, is the penalty value,When the noise estimation figure value less than noiseground truth值的时候,high penalty,When the noise estimate map value is greater than the noiseground truth值的时候,low penalty.（非对称）The purpose is to avoid underestimating the noise value.
2.2、Total Variation Loss Function,The purpose is to limit the smoothness of the noise estimate map（Gradient size to reflect the degree of a smooth）：其中,Calculate the horizontal gradient of the noise estimate map（Noise Estimation Map Vertical Gradient）
2.3、Non-blind noise reduction sub-network loss function：Calculate the pixel-level mean squared error between the output image and the input imageMSE：
The total loss function of the network is the sum of the three parts multiplied by the weights：
RIDNet中,The authors propose a new repair module,Learn features and further enhance the functionality of the network. The author team focuses on features by focusing on dependencies between channels,to rescale channel-level features. 还使用LSC,SSC和SCto bypass low frequency information,So the network can focus on study residual.网络结构如下：
该论文提出了k-sigma变换,使用标定的k和sigma按照k-sigma公式变换后,原始的噪声分布就只和没有噪声的数据x ∗ 有关,因此就可以避免不同增益下噪声不同带来的负担.Its specific principle can refer to：Night scene noise reduction algorithm in your phone-Raw域k-Sigma Transform - 知乎
It is very difficult to obtain pairs of noisy and noise-free images,Therefore, some researchers have used generative adversarial networks（GAN）to generate paired images for training the model,首先,训练GAN网络,Through the noise of the image to study distribution and generate noise samples,Use this to simulate a noisy image in a real scene,解决HRImage lacks correspondenceLR的问题;其次,训练CNN网络,Use the noise blocks sampled in the previous step to construct a paired training dataset,该数据集用于训练CNNto denoise a given image.网络结构如下：
Estimation from noisy images
Since noise images and noise-free images are difficult to obtain,So is it possible to train a good noise model using only noisy images?？Therefore, the researchers proposedNoise2Noise模型,其原理其实很简单：
本来我们做图像降噪,需要输入的噪音图像 x,和 “干净样本” y.例如,x is a path traced rendering image rendered with a few beams,y It is a picture after long-term rendering.那么如果用 y 作为训练目标,生成 y 是个非常费时费力的过程.But in fact, if you think about it,可以用另一次快速渲染生成的另一个噪音图像（它相当于 y + 另一个不同的噪音）作为训练目标（所以叫Noise2Noise）.只要训练样本够多,最终也相当于用 y 作为训练目标.原因是简单的统计学原理.
author[Turned_MZ],Please bring the original link to reprint, thank you.
The sidebar is recommended
- The direction rotation problem of vuejs mobile phone photo upload (solved)
- Camerax preview photo + anti-shake detection
- Android Camera2 tutorial, the fourth chapter, taking pictures
- Gongga Snow Mountain Photography Raiders (2) Lenggacuo, Qumogong, Xueburen Pass
- Vue connects to the high-speed camera to realize the function of taking pictures and uploading on the PC side
- WeChat applet customizes the camera to take pictures, calculate the size, and upload
- Take photos, around the ancient city of...So you are such a tower's door!
- 94. (Cesium House) Cesium Dynamic Monolithization - Oblique Photography (Floor)
- Python controls the mobile phone camera to take pictures (own), and automatically sends the pictures to the mailbox (for learning reference only)
guess what you like
The diffuse wei spider-man heavy plate making secret photo locations of photos have?
94. (Cesium House) Cesium Dynamic Monolithization - Oblique Photography (Floor)
Hong Kong | 12 Influencer Photoshoot Sites, with Detailed Maps and Photography Tips
Astrophotographers capture stunning photos of Saturn as it approaches Earth
Using el-upload to upload is, the mobile phone project cannot add accept, and the mobile phone cannot take pictures after adding
Oblique Photography of SuperMap 3D Project - Optimization of Oblique Photography Data
Java Project: Children's Photography Appointment Website Platform Implemented by SSM
Java project: SSM wedding photo studio photography mall project website
How to take VR panoramic photos with ordinary SLR cameras?
- Extraction of high-precision terrain DEM data based on UAV oblique photography model
- Android CameraX imitates a sweet camera (video, photo, adjustable size, focus, lighting, grid lines), the most complete CameraX tutorial
- The background management system of the photography studio built by Angualr
- [UK Photography] These places have the high-value British style you want
- "Cooling Healing" Get "Islandic Secret" Yunzhong Farm's photo nirvana!
- The photographer invites you to visit the "Wannan Sichuan-Tibet Line", an unmissable four-season scenery~
- java springboot-based photographer appointment photography system ssm
- Actual combat simulation│WeChat JSSDK realizes functions such as custom sharing, mobile phone picture selection and photography, picture audio processing, geographic location, shake and other functions
- Google Camera's anti-sky AI night scene photography technology, RawNeRF, debuts: perfect noise reduction, and can also synthesize 3D perspective
- Astrophotographer Brings Spectacular Moon Details
- Sony announced that it will raise the prices of SLR cameras and other products by about 8% from September 1
- Review | fuyang hangzhou activity of clouds sky camping - time-lapse photography salon moments under the stars!
- Guangzhou shop photography guide, the hiding place of the most beautiful old days
- Head of Meta virtual reality platform Horizon leaves after Zuckerberg Metaverse selfie gets ridiculed
- php photography picture album sharing website management system mysql
- Automatic keying change background software-cutting photo software
- Weihai Time-lapse photography "Looking for Liugong Island"
- Bashang WuLanBu series autumn photography tips
- Bashang Baima Culture Theme Photography Guide
- The third day animal behavior photo guide of "Genshin Impact"
- Cesium: Loading OSGB Oblique Photography 3D Models
- Aerial Oblique Photography Web 3D GIS Digital Twin Smart Thermal Power Plant
- Taking pictures, visiting the ancient city... It turns out that you are such a Thapae Gate!
- Smart Community Management Cloud Platform - Digital Twin Large-screen 3D Visualization - Easy Shooting Mini Program - Inclined Photography Layered Accounts
- A maid dressed as Deadpool appeared at Tokyo Comic Con!The coquettish shape attracts countless audiences to watch and take pictures!
- 2019 Seda Travel Guide
- It takes two hours to take pictures and retouch for 5 seconds. This plugin is amazing.
- "Twisted Car Boy" Zhang Liang's enrollment: plans to open a photography studio during college
- How to convert a photographed image into a clearer image?
- Feel the sweetness and beauty of New Zealand through the lens of this celebrity photographer
- Ningbo Ningbo where to go on weekends?These three super niche places are beautiful to take pictures!
- springboot photographic equipment equipment rental system java
- Which translation photo software is better?Photo software recommended!
- Xishuangbanna Winter is coming, the little sisters who came to Xishuangbanna to take pictures, take a walk in the guide to avoid pits
- About cesium loading a large amount of oblique photography data is slow to load
- android camera2 take pictures
- Hefei goes beyond the big collection of photo shoots and wins the ticket circle
- Ningbo Ningbo Mid-Autumn Festival Tour, take pictures and break into the autumn fairy tale world, don't go to wait another year!
- Multi-functional beauty face changing and photo text recognition application based on Android+JavaWEB