pytorch图像加噪

skimage中random_noise使用的一些问题：

random_noise(image, mode='gaussian', seed=None, clip=True, **kwargs)
    Function to add random noise of various types to a floating-point image.
    
    Parameters
    ----------
    image : ndarray 
        Input image data. Will be converted to float. #输入为ndarry，输出为float
    mode : str, optional
        One of the following strings, selecting the type of noise to add:
#mode名字：
        - 'gaussian'  Gaussian-distributed additive noise.
        - 'localvar'  Gaussian-distributed additive noise, with specified
                      local variance at each point of `image`.
        - 'poisson'   Poisson-distributed noise generated from the data.
        - 'salt'      Replaces random pixels with 1.
        - 'pepper'    Replaces random pixels with 0 (for unsigned images) or
                      -1 (for signed images).
        - 's&p'       Replaces random pixels with either 1 or `low_val`, where
                      `low_val` is 0 for unsigned images or -1 for signed
                      images.
        - 'speckle'   Multiplicative noise using out = image + n*image, where
                      n is uniform noise with specified mean & variance.
    seed : int, optional
        If provided, this will set the random seed before generating noise,
        for valid pseudo-random comparisons.

#范围限定（对我来说，用默认clip(True)即可，我认为） 
   clip : bool, optional
        If True (default), the output will be clipped after noise applied
        for modes `'speckle'`, `'poisson'`, and `'gaussian'`. This is
        needed to maintain the proper image data range. If False, clipping
        is not applied, and the output may extend beyond the range [-1, 1].
    mean : float, optional
        Mean of random distribution. Used in 'gaussian' and 'speckle'.
        Default : 0.

#随机添加噪声的强度（方差，guassian, 范围[0.1],默认0.01）
    var : float, optional
        Variance of random distribution. Used in 'gaussian' and 'speckle'.
        Note: variance = (standard deviation) ** 2. Default : 0.01
    local_vars : ndarray, optional
        Array of positive floats, same shape as `image`, defining the local
        variance at every image point. Used in 'localvar'.

#随机添加噪声的强度（像素点个数， 椒盐， 范围[0.1], 默认0.05）
    amount : float, optional
        Proportion of image pixels to replace with noise on range [0, 1].
        Used in 'salt', 'pepper', and 'salt & pepper'. Default : 0.05
    salt_vs_pepper : float, optional
        Proportion of salt vs. pepper noise for 's&p' on range [0, 1].
        Higher values represent more salt. Default : 0.5 (equal amounts)
    
    Returns
    -------
    out : ndarray
        Output floating-point image data on range [0, 1] or [-1, 1] if the
        input `image` was unsigned or signed, respectively.

自己修改的代码：

def read_image(data_path):  # 读取二进制文件并返回归一化后的图片
    with open(data_path, 'rb') as f:
        data1 = np.fromfile(f, dtype=np.uint8)
        image = np.reshape(data1, (-1, 3, 96, 96))
        images = np.transpose(image, (0, 3, 2, 1))#转换通道.转成RGB通道
    return images/255.0

data_path = "/content/drive/MyDrive/train_X.bin"  # 数据集路径，r表示防止转义
images = read_image(data_path)

def gaussian_noise(images, sigma):
    """sigma: 噪声标准差""" #默认0.01,范围[0,1]
    sigma2 = sigma**2 #/ (255 ** 2)   # 噪声方差
    images_noisy = np.zeros_like(images)
    for ii in range(images.shape[0]):
        image = images[ii]
        # 使用skimage中的函数增加噪音
        noise_im = random_noise(image, mode="gaussian", var=sigma2, clip=True)
        images_noisy[ii] = noise_im
    return images_noisy

images_noise = gaussian_noise(images, 0.1)
print("image_noise:", images_noise.min(), "~", images_noise.max())

def salt_speckle_noise(images, ratio):
    """amount: 噪声数量占比""" #默认0.05,范围[0,1]

    images_noisy = np.zeros_like(images)
    for ii in range(images.shape[0]):
        image = images[ii]
        # 使用skimage中的函数增加噪音
        noise_im = random_noise(image, mode="s&p", amount=ratio) #salt_vs_pepper椒、盐噪声比默认0.5，范围[0.1]
        images_noisy[ii] = noise_im
    return images_noisy

images_noise = salt_speckle_noise(images, 0.05）
print("image_noise:", images_noise.min(), "~", images_noise.max())

关于加噪的其他注意问题：

1.使用 skimage.util.random_noise(img, mode="salt")会导致使用opencv去降噪的函数会提示无法支持的格式 ！ 

问题的所在

skimage.util.random_noise（）

out : ndarray

Output floating-point image data on range [0, 1] or [-1, 1] if the
input `image` was unsigned or signed, respectively.

该加噪函数（方法）将图像转化[0,1]或者[-1，1],而且数据类型变成了float64类型，而opencv处理的图像格式一般是uint8类型的，因此还需要进行转化。

首先需将[0，1]转化为0-255,然后将浮点float64转为uint8的。因为代码跑在服务器上没有图像界面，因此需要用visdom做图片的显示。附代码：

import cv2
import visdom
import numpy as np
import matplotlib.pyplot as plt
from skimage.util.dtype import convert
import skimage.util
 
viz = visdom.Visdom(env='img')
img = cv2.imread('qiao.jpg', 1)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
 
print(img.dtype)
#img=img.transpose(1,0,2)
# 使用一下函数进行加噪声会导致数据类型从uint8转为float64，而去噪的 函数只能处理8位的uint类型的图片
noise_img = skimage.util.random_noise(img, mode="salt")
# 将加了噪音的[0,1]之间转化为0-255
noise_img = noise_img*255
 
plt.figure(0)
noise_img = noise_img.astype(np.uint8)
viz.image(noise_img.transpose(2, 0, 1))
# denoise = cv2.GaussianBlur(noise_img, (5, 5), 1)
denoise = cv2.GaussianBlur(noise_img, ksize=(3, 3), sigmaX=0)
viz.image(denoise.transpose(2, 0, 1))
denoise_mean = cv2.medianBlur(noise_img, 5)
viz.image(denoise_mean.transpose(2, 0, 1))
denoise_mean = cv2.fastNlMeansDenoising(noise_img,h=30)
viz.text("均值滤波结果")
viz.image(denoise_mean.transpose(2, 0, 1))