训练模型需要的数据文件有：

MNIST_data文件夹下的mnist_train、mnist_test、noisy_train、noisy_test。train文件夹下60000个图片，test下10000个图片

noisy_train、noisy_test下的图片加了椒盐噪声与原图序号对应

离线测试需要的数据文件有：

MNIST_data文件夹下的my_model.hdf5、my_test。my_test文件夹下要有一层嵌套文件夹并放测试图片

数据集准备参考：

https://www.cnblogs.com/dzzy/p/10824072.html

训练：

import os
import glob
from PIL import Image
import numpy as np
from warnings import simplefilter
simplefilter(action='ignore', category=FutureWarning)
import matplotlib.pyplot as plt
from keras.models import Model
from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
from keras.preprocessing.image import ImageDataGenerator,img_to_array, load_img
from keras.callbacks import TensorBoard , ModelCheckpoint
print("_________________________keras start_____________________________")


base_dir = 'MNIST_data' #基准目录
#os._exit(0)
Datapath = os.path.join(base_dir,'mnist_train/*.png') #train目录
x_train = np.zeros((60000, 28, 28))
x_train = np.reshape(x_train, (60000, 28, 28, 1))
i = 0
for imageFile in glob.glob(Datapath ):
    # 打开图像并转化为数字矩阵
    img = np.array(Image.open(imageFile))
    img = np.reshape(img, (1, 28, 28, 1))
    img = img.astype('float32') / 255.
    x_train[i] = img
    i += 1

Datapath = os.path.join(base_dir,'mnist_test/*.png') #test目录
x_test = np.zeros((10000, 28, 28))
x_test = np.reshape(x_test, (10000, 28, 28, 1))
i = 0
for imageFile in glob.glob(Datapath ):
    # 打开图像并转化为数字矩阵
    img = np.array(Image.open(imageFile))
    img = np.reshape(img, (1, 28, 28, 1))
    img = img.astype('float32') / 255.
    x_test[i] = img
    i += 1

print( x_train.shape)
print( x_test.shape)

Datapath = os.path.join(base_dir,'noisy_train/*.png') #test目录
x_train_noisy = np.zeros(x_train.shape)
i = 0
for imageFile in glob.glob(Datapath ):
    # 打开图像并转化为数字矩阵
    img = np.array(Image.open(imageFile))
    img = np.reshape(img, (1, 28, 28, 1))
    img = img.astype('float32') / 255.
    x_train_noisy[i] = img
    i += 1

Datapath = os.path.join(base_dir,'noisy_test/*.png') #test目录
x_test_noisy = np.zeros(x_test.shape)
i = 0
for imageFile in glob.glob(Datapath ):
    # 打开图像并转化为数字矩阵
    img = np.array(Image.open(imageFile))
    img = np.reshape(img, (1, 28, 28, 1))
    img = img.astype('float32') / 255.
    x_test_noisy[i] = img
    i += 1

print( x_train_noisy.shape)
print( x_test_noisy.shape)

'''
plt.figure(figsize=(20, 4))
plt.subplot(4, 4, 1)
plt.imshow(x_train[0].reshape(28, 28))
plt.subplot(4, 4, 2)
plt.imshow(x_train_noisy[0].reshape(28, 28))
plt.subplot(4, 4, 3)
plt.imshow(x_train[1].reshape(28, 28))
plt.subplot(4, 4, 4)
plt.imshow(x_train_noisy[1].reshape(28, 28))
plt.show()
#os._exit(0)
'''

"""
搭建模型
"""
input_img = Input(shape=(28, 28, 1))  # adapt this if using `channels_first` image data format

x = Conv2D(32, (3, 3), activation='relu', padding='same')(input_img) #relu激活函数
x = MaxPooling2D((2, 2), padding='same')(x)
x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling2D((2, 2), padding='same')(x)

# at this point the representation is (7, 7, 32)

x = Conv2D(32, (3, 3), activation='relu', padding='same')(encoded)
x = UpSampling2D((2, 2))(x)
x = Conv2D(32, (3, 3), activation='relu', padding='same')(x)
x = UpSampling2D((2, 2))(x)
decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)

autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')

file_path="MNIST_data/weights-improvement-{epoch:02d}-{val_loss:.2f}.hdf5"
tensorboard = TensorBoard(log_dir='/tmp/tb', histogram_freq=0, write_graph=False)
checkpoint = ModelCheckpoint(filepath=file_path,verbose=1,monitor='val_loss', save_weights_only=False,mode='auto' ,save_best_only=True,period=1)
autoencoder.fit(x_train_noisy, x_train,
                epochs=100,
                batch_size=128,
                shuffle=True,
                validation_data=(x_test_noisy, x_test),
                callbacks=[checkpoint,tensorboard])
                
#展示结果                
n = 10
plt.figure(figsize=(20, 4))
for i in range(n):
#noisy data
    ax = plt.subplot(3, n, i+1)
    plt.imshow(x_test_noisy[i].reshape(28, 28))
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
#predict
    ax = plt.subplot(3, n, i+1+n)
    decoded_img = autoencoder.predict(x_test_noisy)
    plt.imshow(decoded_img[i].reshape(28, 28))
    plt.gray()
    ax.get_yaxis().set_visible(False)
    ax.get_xaxis().set_visible(False)
#original
    ax = plt.subplot(3, n, i+1+2*n)
    plt.imshow(x_test[i].reshape(28, 28))
    plt.gray()
    ax.get_yaxis().set_visible(False)
    ax.get_xaxis().set_visible(False)
plt.show()

测试：（同  https://www.cnblogs.com/dzzy/p/11387645.html）

import os
import numpy as np
from warnings import simplefilter
simplefilter(action='ignore', category=FutureWarning)
import matplotlib.pyplot as plt
from keras.models import Model,Sequential,load_model
from keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
from keras.preprocessing.image import ImageDataGenerator,img_to_array, load_img
from keras.callbacks import TensorBoard , ModelCheckpoint
print("_________________________keras start_____________________________")
pic_num = 3
base_dir = 'MNIST_data' #基准目录
train_dir = os.path.join(base_dir,'my_test') #train目录
validation_dir="".join(train_dir)
test_datagen = ImageDataGenerator(rescale= 1./255)
validation_generator  = test_datagen.flow_from_directory(validation_dir,
                                                    target_size = (28,28),
                                                    color_mode = "grayscale",
                                                    batch_size = pic_num,
                                                    class_mode =  "categorical")#利用test_datagen.flow_from_directory（图像地址，目标size，批量数目，标签分类情况）
for x_train,batch_labels in validation_generator:
    break
x_train = np.reshape(x_train, (len(x_train), 28, 28, 1))
y_train = x_train

# create model
model = load_model('MNIST_data/my_model.hdf5')
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
print("Created model and loaded weights from file")

# estimate accuracy on whole dataset using loaded weights
y_train=model.predict(x_train)
# 评价训练出的网络
#loss, accuracy = model.evaluate(x_train, y_train)
#print('test loss: ', loss)
#print('test accuracy: ', accuracy)

n = pic_num
for i in range(n):
    ax = plt.subplot(2, n, i+1)
    plt.imshow(x_train[i].reshape(28, 28))
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
    ax = plt.subplot(2, n, i+1+n)
    plt.imshow(y_train[i].reshape(28, 28))
    plt.gray()
    ax.get_xaxis().set_visible(False)
    ax.get_yaxis().set_visible(False)
plt.show()