python检测图片噪声（噪点噪声、雪花噪声、条纹噪声）

首先是噪声的大体分类：

噪点噪声：又称脉冲噪声、椒盐噪声

雪花噪声：又称高斯噪声

条纹噪声：

细节图如下所示（图像来源，论文http://www.doc88.com/p-2572496212147.html）

分析完这些噪声的大致分布情况之后

首先需要作出这些噪声图（原型来自https://www.jb51.net/article/162073.htm）

import cv2
from PIL import Image
from PIL import ImageChops
import numpy as np
import time
import pytesseract
import warnings
import math
import random
 
def sp_noise(image,prob=0.05):
  '''
  添加椒盐噪声
  prob:噪声比例 
  '''
  output = np.zeros(image.shape,np.uint8)
  thres = 1 - prob 
  for i in range(image.shape[0]):
    for j in range(image.shape[1]):
      rdn = random.random()
      if rdn < prob:
        output[i][j] = 0
      elif rdn > thres:
        output[i][j] = 255
      else:
        output[i][j] = image[i][j]
  #return output
  #print(output)
  cv2.imshow("img",output)
  cv2.waitKey(0)
  cv2.imwrite("noise_check/img.jpg",output)
 
 
def gasuss_noise(image, mean=0.2, var=0.005):
  ''' 
    添加高斯噪声
    mean : 均值 
    var : 方差
  '''
  image = np.array(image/255, dtype=float)
  noise = np.random.normal(mean, var ** 0.5, image.shape)
  out = image + noise
  if out.min() < 0:
    low_clip = -1.
  else:
    low_clip = 0.
  out = np.clip(out, low_clip, 1.0)
  out = np.uint8(out*255)
  cv2.imshow("gasuss", out)
  #return out
  cv2.waitKey(0)
  cv2.imwrite("noise_check/img.jpg",out)

#sp_noise(cv2.imread("noise_check/5.jpg",cv2.IMREAD_COLOR))
gasuss_noise(cv2.imread("noise_check/5.jpg",cv2.IMREAD_COLOR))

检测噪点和雪花的代码如下（均方误差法，思路来源https://blog.csdn.net/twinkle_star1314/article/details/74858253）

import cv2
from PIL import Image
from PIL import ImageChops
import numpy as np
import time
import pytesseract
import warnings

warnings.filterwarnings("ignore")
demo=Image.open("noise_check//1.jpg")
im=np.array(demo.convert('L'))#灰度化矩阵
print(im.shape)
print(im.dtype)
#print(im)
height=im.shape[0]#尺寸
width=im.shape[1]
varlist=[]
for i in range(height):
    for j in range(width):
        for k in range(16):
            if im[i][j]>=k*16 and im[i][j]<(k+1)*16:#16级量化
                im[i][j]=8*(k*2+1)
                break
for i in range(0,height-height%3,3):
    for j in range(0,width-width%3,3):
        x=(im[i][j]+im[i+1][j]+im[i+2][j]+im[i][j+1]+im[i+1][j+1]+im[i+2][j+1]+im[i][j+2]+im[i+1][j+2]+im[i+2][j+2])/9
        x2=(pow(im[i][j],2)+pow(im[i+1][j],2)+pow(im[i+2][j],2)+pow(im[i][j+1],2)+pow(im[i+1][j+1],2)+pow(im[i+2][j+1],2)+pow(im[i][j+2],2)+pow(im[i+1][j+2],2)+pow(im[i+2][j+2],2))/9
        var=x2-pow(x,2)
        varlist.append(round(var,3))#子窗口的方差值3x3
print(im)
#print(varlist)
T=round(sum(varlist)/len(varlist),3)#保留3位小数
print(T)

检测噪点和雪花的方法如下（FFT法）

from PIL import Image
import numpy as np
import math

T=50#阈值设定，大于T则判定偏离xy轴过多
 
#复数类
class complex:
    def __init__(self):
        self.real = 0.0
        self.image = 0.0
 
#复数乘法
def mul_ee(complex0, complex1):
    complex_ret = complex()
    complex_ret.real = complex0.real * complex1.real - complex0.image * complex1.image
    complex_ret.image = complex0.real * complex1.image + complex0.image * complex1.real
    return complex_ret
 
#复数加法
def add_ee(complex0, complex1):
    complex_ret = complex()
    complex_ret.real = complex0.real + complex1.real
    complex_ret.image = complex0.image + complex1.image
    return complex_ret
 
#复数减法
def sub_ee(complex0, complex1):
    complex_ret = complex()
    complex_ret.real = complex0.real - complex1.real
    complex_ret.image = complex0.image - complex1.image
    return complex_ret
 
#对输入数据进行倒序排列
def forward_input_data(input_data, num):    
    j = num //2
    for i in range(1, num - 2):        
        if(i < j):
            complex_tmp = input_data[i]
            input_data[i] = input_data[j]
            input_data[j] = complex_tmp
            #print "forward x[%d] <==> x[%d]" % (i, j)
        k = num // 2
        while (j >= k):
            j = j - k
            k = k // 2
        j = j + k
 
#实现1D FFT
def fft_1d(in_data, num):
    PI = 3.1415926
    forward_input_data(in_data, num) #倒序输入数据    
 
    #计算蝶形级数，也就是迭代次数
    M = 1 #num = 2^m
    tmp = num // 2;
    while (tmp != 1):
        M = M + 1
        tmp = tmp // 2
    #print "FFT level：%d" % M
 
    complex_ret = complex()
    for L in range(1, M + 1):
        B = int(math.pow(2, L -1)) #B为指数函数返回值，为float，需要转换integer
        for J in range(0, B):
            P = math.pow(2, M - L) * J            
            for K in range(J, num, int(math.pow(2, L))):
                #print "L:%d B:%d, J:%d, K:%d, P:%f" % (L, B, J, K, P)
                complex_ret.real = math.cos((2 * PI / num) *  P)
                complex_ret.image = -math.sin((2 * PI / num) * P)
                complex_mul = mul_ee(complex_ret, in_data[K + B])
                complex_add = add_ee(in_data[K], complex_mul)
                complex_sub = sub_ee(in_data[K], complex_mul)
                in_data[K] = complex_add
                in_data[K + B] = complex_sub
                #print "A[%d] real: %f, image: %f" % (K, in_data[K].real, in_data[K].image)
               # print "A[%d] real: %f, image: %f" % (K + B, in_data[K + B].real, in_data[K + B].image)
 
def test_fft_1d(in_data):
    #in_data = [2,3,4,5,7,9,10,11,100,12,14,11,56,12,67,12] #待测试的x点元素
    k=1
    while(1):
        if len(in_data)>pow(2,k) and len(in_data)<=pow(2,k+1):#不足的补0
            #fftlen=pow(2,k+1)
            #in_data.extend([0 for i in range(pow(2,k+1)-len(in_data))])
            fftlen=pow(2,k)
            break
        k+=1
    #变量data为长度为x、元素为complex类实例的list，用于存储输入数据
    data = [(complex()) for i in range(len(in_data))]
    #将8个测试点转换为complex类的形式，存储在变量data中
    for i in range(len(in_data)):
        data[i].real = in_data[i]
        data[i].image = 0.0
         
    ##输出FFT需要处理的数据
    #print("The input data:")
    #for i in range(len(in_data)):
    #    print("x[%d] real: %f, image: %f" % (i, data[i].real, data[i].image))
          
    fft_1d(data, fftlen)
 
    ##输出经过FFT处理后的结果
    #print("The output data:")
    #for i in range(len(in_data)):
       # print("X[%d] real: %f, image: %f" % (i, data[i].real, data[i].image))

    Tnum=0
    for i in range(len(in_data)):#虚实值都大于T的才叫偏离
        if abs(data[i].real)>T and abs(data[i].image)>T:
            Tnum+=1
    print(Tnum)
    print(str(round(Tnum/len(in_data),4)*100)+"%")
    
#test the 1d fft
#in_data=[2,3,4,5,7,9,10,11]
demo=Image.open("noise_check//5.jpg")
im=np.array(demo.convert('L'))#灰度化矩阵
in_data=[]
for item in im:
    in_data.extend(item)
test_fft_1d(in_data)

以下为原图、均方误差法结果、FFT法结果

可以看出FFT法比均方误差法要准确，虽然时间上也更长···

对于正常图片，这个FFT百分比一般不超过94%

对于噪声较小的也能在这个数值上体现出来：

要是更小的噪声的话···

可能准确率就不行了···

条纹噪声的检测和上述不同

不知道如何才能生成条纹噪声···

按照上面那个论文的思路倒是将代码写了出来，还未经过测试所以正确率不能保证

from PIL import Image
import numpy as np
import warnings

T1=100#阈值1，通道行差
T2=1000#阈值2，A通道差绝对和
T3=1000#阈值3，AB通道绝对和
#算法来源，论文http://www.doc88.com/p-2572496212147.html
warnings.filterwarnings("ignore")
demo=Image.open("noise_check//21.jpg")
im=np.array(demo.convert('L'))#灰度化矩阵
print(im.shape)
print(im.dtype)
r,g,b=demo.split()
#gm=demo.convert('L')
#plt.subplot(2,2,1)
#plt.imshow(gm,cmap='gray'),plt.axis('off')
#plt.subplot(2,2,2)
#plt.imshow(r,cmap='gray'),plt.axis('off')
#plt.subplot(2,2,3)
#plt.imshow(g,cmap='gray'),plt.axis('off')
#plt.subplot(2,2,4)
#plt.imshow(b,cmap='gray'),plt.axis('off')
#plt.show()
rm=np.array(r)
gm=np.array(g)
bm=np.array(b)
height=im.shape[0]#尺寸
width=im.shape[1]
midimg=[im,rm,gm,bm]
count=0
for i in range(4):
    mid=midimg[i]
    n=0
    while(1):
        if n+3>=height:break
        for j in range(6,width-6,1):
            grA=mid[n][j]
            grB=mid[n+3][j]
            if abs(grA-grB)>T1:
                L1=0
                L2=0
                for k in range(j-6,j+6,1):
                    L1+=abs(mid[n][k]-grA)
                    L2+=abs(mid[n][k]-mid[n+3][k])
                    #print(L1)
                    #print(L2)
                    #print("-----")
                    if L1<T2 and L2>T3:
                        count+=1
        n+=10
#print(count)
sum=(height-3)*4*(width-12)
#print(count/sum)
res=round(count/sum,5)#保留3位小数
print(str(res*100)+"%")