tensorflow下的图片标准化函数per_image_standardization用法

import tensorflow as tf
import matplotlib.image as img
import matplotlib.pyplot as plt
import numpy as np
sess = tf.InteractiveSession()
image = img.imread("D:/Documents/Pictures/logo7.jpg")
shape = tf.shape(image).eval()
h,w = shape[0],shape[1]
standardization_image = tf.image.per_image_standardization(image)#标准化

fig = plt.figure()
fig1 = plt.figure()
ax = fig.add_subplot(111)
ax.set_title("orginal image")
ax.imshow(image)
ax1 = fig1.add_subplot(311)
ax1.set_title("original hist")
ax1.hist(sess.run(tf.reshape(image,[h*w,-1])))
ax1 = fig1.add_subplot(313)
ax1.set_title("standardization hist")
ax1.hist(sess.run(tf.reshape(standardization_image,[h*w,-1])))
plt.ion()
plt.show()

saver = tf.train.Saver()#声明saver用来保存模型
with tf.Session() as sess:
 for i in range(train_step):
 #.....训练操作
 if i%100 == 0 && i!= 0:#每间隔训练100次存储一个模型，默认最多能存5个，如果超过5个先将序号小的覆盖掉
  saver.save(sess,str(i)+"_"+"model.ckpt",global_step=i)

saver = tf.train.Saver()
with tf.Session() as sess:
 saver.restore(sess, "step_1497batch_64model.ckpt-1497")#加载对应的参数

def read_data(tfrecords_file, batch_size, image_size):#读取tfrecords文件
 filename_queue = tf.train.string_input_producer([tfrecords_file])
 reader = tf.TFRecordReader()
 _, serialized_example = reader.read(filename_queue)
 
 img_features = tf.parse_single_example(
  serialized_example,
 features={
 "label": tf.FixedLenFeature([], tf.int64),
 "image_raw": tf.FixedLenFeature([], tf.string),
  })
 image = tf.decode_raw(img_features["image_raw"], tf.float32)
 min_after_dequeue = 1000
 image = tf.reshape(image, [image_size, image_size,1])
 image = tf.image.resize_images(image, (32,32),method=3)#缩放成32×32
 image = tf.image.per_image_standardization(image)#图像标准化
 label = tf.cast(img_features["label"], tf.int32)  
 capacity = min_after_dequeue + 3 * batch_size  
 image_batch, label_batch = tf.train.shuffle_batch([image, label],
       min_after_dequeue = min_after_dequeue)
 return image_batch, tf.one_hot(label_batch,6)#返回的标签经过one_hot编码

#将得到的图像数据与标签都是tensor哦，不能输出的！
read_image_batch,read_label_batch = read_data("train_data\tfrecord\TrainC6_95972.tfrecords",batch_size,120)

image = tf.reshape(image, [image_size, image_size,1])#与opencv的reshape结果一样
image = tf.image.resize_images(image, (32,32),method=3)#缩放成32×32，与opencv的resize结果一样，插值方法要选择三次立方插值
image = tf.image.per_image_standardization(image)#图像标准化

with tf.name_scope("inputs"):
 img_raw = tf.placeholder(dtype=tf.float32, shape=[1,120, 120, 1], name="input")#输入数据
 keep_prob = tf.placeholder(tf.float32,name="keep_prob")

with tf.name_scope("preprocess"):#图中的预处理函数，当成节点顺序调用
 img_120 = tf.reshape(img_raw, [120, 120,1])
 img_norm = tf.cast(img_120, "float32") / 256
 img_32 = tf.image.resize_images(img_norm, (32,32),method=3)
 img_std = tf.image.per_image_standardization(img_32)
 img = tf.reshape(img_std, [1,32, 32,1])

with tf.name_scope("output"):#图像塞到网络中
 output = MyNet(img,keep_prob,n_cls)

ans = tf.argmax(tf.nn.softmax(output),1)#计算模型得到的结果

init = tf.global_variables_initializer()

saver = tf.train.Saver()


if __name__ == "__main__":

 with tf.Session() as sess:

 sess.run(init)
 saver.restore(sess, "step_1497batch_64model.ckpt-1497")#效果更好
 index = 0
 path = "buffer\"
 
 while True:
  f = path + str(index)+".jpg"#从0.jpg、1.jpg、2.jpg.....一直读
  if os.path.exists(f):
  cv_img = cv.imread(f,0)
  cv_img = OneImgPrepro(cv_img)
  cv_img = cv_img.reshape(1,cv_img.shape[0],cv_img.shape[1],1)#需要reshape成placeholder可接收型
  clas = ans.eval(feed_dict={img_raw:cv_img,keep_prob:1})#feed的速度快！

  print(clas)#输出分类
  
  index += 1