使用深度学习进行图像处理

使用深度神经网络执行图像处理任务，例如去除图像噪声和执行图像到图像的转换（需要 Deep Learning Toolbox™）

深度学习使用神经网络直接从数据中学习有用的特征表示。例如，您可以使用预训练神经网络来识别和去除图像中的噪声等项。

A noisy image is passed to a trained denoising network, which returns a denoised image.

函数

创建数据存储以进行图像预处理

`augmentedImageDatastore`	变换批量以增强图像数据
`blockedImageDatastore`	Datastore for use with blocks from `blockedImage` objects (自 R2021a 起)
`denoisingImageDatastore`	Denoising image datastore
`imageDatastore`	图像数据的数据存储
`randomPatchExtractionDatastore`	Datastore for extracting random 2-D or 3-D random patches from images or pixel label images
`transform`	变换数据存储
`combine`	合并来自多个数据存储的数据

增强图像

`jitterColorHSV`	Randomly alter color of pixels (自 R2019b 起)
`randomWindow2d`	Randomly select rectangular region in image (自 R2021a 起)
`randomCropWindow3d`	Create randomized cuboidal cropping window (自 R2019b 起)
`centerCropWindow2d`	Create rectangular center cropping window (自 R2019b 起)
`centerCropWindow3d`	Create cuboidal center cropping window (自 R2019b 起)
`Rectangle`	Spatial extents of 2-D rectangular region (自 R2019b 起)
`Cuboid`	Spatial extents of 3-D cuboidal region (自 R2019b 起)
`randomAffine2d`	Create randomized 2-D affine transformation (自 R2019b 起)
`randomAffine3d`	Create randomized 3-D affine transformation (自 R2019b 起)
`affineOutputView`	Create output view for warping images (自 R2019b 起)
`imerase`	Remove image pixels within rectangular region of interest (自 R2021a 起)

调整深度学习输入的大小和形状

`resize2dLayer`	2-D resize layer (自 R2020b 起)
`resize3dLayer`	3-D resize layer (自 R2020b 起)
`dlresize`	Resize spatial dimensions of `dlarray` object (自 R2020b 起)
`DepthToSpace2DLayer`	Depth to space layer (自 R2021a 起)
`SpaceToDepthLayer`	Space to depth layer (自 R2020b 起)
`depthToSpace`	Rearrange `dlarray` data from depth dimension into spatial blocks (自 R2021a 起)
`spaceToDepth`	Rearrange spatial blocks of `dlarray` data along depth dimension (自 R2021a 起)

创建深度学习网络

`encoderDecoderNetwork`	Create encoder-decoder network (自 R2021a 起)
`blockedNetwork`	Create network with repeating block structure (自 R2021a 起)
`pretrainedEncoderNetwork`	Create encoder network from pretrained network (自 R2021a 起)
`cycleGANGenerator`	Create CycleGAN generator network for image-to-image translation (自 R2021a 起)
`patchGANDiscriminator`	Create PatchGAN discriminator network (自 R2021a 起)
`pix2pixHDGlobalGenerator`	Create pix2pixHD global generator network (自 R2021a 起)
`addPix2PixHDLocalEnhancer`	Add local enhancer network to pix2pixHD generator network (自 R2021a 起)
`unitGenerator`	Create unsupervised image-to-image translation (UNIT) generator network (自 R2021a 起)
`unitPredict`	Perform inference using unsupervised image-to-image translation (UNIT) network (自 R2021a 起)

图像去噪

`denoiseImage`	Denoise image using deep neural network
`denoisingNetwork`	获取图像去噪网络
`dnCNNLayers`	Get denoising convolutional neural network layers

主题

预处理图像数据以进行深度学习

Get Started with Image Preprocessing and Augmentation for Deep Learning

Preprocess data for deep learning applications with deterministic operations such as resizing, or augment training data with randomized operations such as random cropping.
Datastores for Deep Learning (Deep Learning Toolbox)
Learn how to use datastores in deep learning applications.
为图像到图像的回归准备数据存储 (Deep Learning Toolbox)
此示例说明如何准备数据存储，以便使用 ImageDatastore 的 transform 和 combine 函数来训练图像到图像的回归网络。

创建用于图像处理应用的神经网络

Train and Apply Denoising Neural Networks
Use a pretrained neural network to remove Gaussian noise from a grayscale image, or train your own network using predefined layers.
Create Modular Neural Networks
You can create and customize deep learning networks that follow a modular pattern with repeating groups of layers, such as U-Net and cycleGAN.
Get Started with GANs for Image-to-Image Translation
Transfer styles and characteristics from one set of images to the scene content of other images by using generative adversarial networks (GANs).
预训练的深度神经网络 (Deep Learning Toolbox)
了解如何下载和使用预训练的卷积神经网络进行分类、迁移学习和特征提取。
深度学习层列表 (Deep Learning Toolbox)
探索 MATLAB^® 中的所有深度学习层。

在 MATLAB 中进行深度学习

在 MATLAB 中进行深度学习 (Deep Learning Toolbox)
通过使用卷积神经网络进行分类和回归来探索 MATLAB 的深度学习能力，包括预训练网络和迁移学习，以及在 GPU、CPU、集群和云上进行训练。
Semantic Segmentation Using Deep Learning (Computer Vision Toolbox)
This example shows how to segment an image using a semantic segmentation network.

精选示例

使用深度学习增加图像分辨率

此示例说明如何使用超深超分辨率 (VDSR) 神经网络从低分辨率图像创建高分辨率图像。

打开实时脚本

Unsupervised Day-to-Dusk Image Translation Using UNIT

Translate images between daytime and dusk lighting conditions using an unsupervised image-to-image translation network (UNIT).

打开实时脚本

Detect Image Anomalies Using Explainable FCDD Network

Use an anomaly detector to distinguish between normal pills and pills with anomalous chips or contamination.

打开实时脚本

Classify Defects on Wafer Maps Using Deep Learning

Classify eight types of manufacturing defects on wafer maps using a simple convolutional neural network (CNN).

打开实时脚本

Detect Image Anomalies Using Pretrained ResNet-18 Feature Embeddings

Train a similarity-based anomaly detector using one-class learning of feature embeddings extracted from a pretrained ResNet-18 convolutional neural network.

打开实时脚本