如何用python实现图像的一维高斯滤波器

如何用python实现图像的一维高斯滤波器

现在把卷积模板中的值换一下，不是全1了，换成一组符合高斯分布的数值放在模板里面，比如这时中间的数值最大，往两边走越来越小，构造一个小的高斯包。实现的函数为cv2.GaussianBlur()。对于高斯模板，我们需要制定的是高斯核的高和宽（奇数），沿x与y方向的标准差(如果只给x，y=x，如果都给0，那么函数会自己计算)。高斯核可以有效的出去图像的高斯噪声。当然也可以自己构造高斯核，相关函数：cv2.GaussianKernel().

import cv2

import numpy as np

import matplotlib.pyplot as plt

img = cv2.imread(‘flower.jpg‘,0) #直接读为灰度图像

for i in range(2000): #添加点噪声

temp_x = np.random.randint(0,img.shape[0])

temp_y = np.random.randint(0,img.shape[1])

img[temp_x][temp_y] = 255

blur = cv2.GaussianBlur(img,(5,5),0)

plt.subplot(1,2,1),plt.imshow(img,‘gray‘)#默认彩色，另一种彩色bgr

plt.subplot(1,2,2),plt.imshow(blur,‘gray‘)

import tensorflow as tf

import cv2

import os

import numpy as np

os.environ["CUDA_VISIBLE_DEVICES"] = "3"

path="/home/ta/Desktop/gs_for_test/1.bmp"

ori=cv2.imread(path, cv2.IMREAD_GRAYSCALE)

img=tf.convert_to_tensor(ori)

noise=tf.random_normal(shape=img.shape, mean=0, stddev=50)

drop_noise=tf.nn.dropout(noise, 0.01)

final=tf.add(img, tf.cast(drop_noise, dtype=tf.uint8))

cv2.namedWindow('img',cv2.WINDOW_KEEPRATIO)

cv2.namedWindow('final',cv2.WINDOW_KEEPRATIO)

with tf.Session() as sess:

      fimg=sess.run(final)

      img__=sess.run(img)

      cv2.imshow("final", fimg)

      cv2.imshow("img", img__)

       cv2.waitKey()

import CV2

import copy

import numpy as np

import random

使用的是pycharm

因为最近看了《银翼杀手2049》，里面Joi实在是太好看了所以原图像就用Joi了

要求是灰度图像，所以第一步先把图像转化成灰度图像

# 读入原始图像

img = CV2.imread('joi.jpg')

# 灰度化处理

gray = CV2.cvtColor(img, CV2.COLOR_BGR2GRAY)

CV2.imwrite('img.png', gray)

第一个任务是利用分段函数增强灰度对比，我自己随便写了个函数大致是这样的

def chng(a):

if a <255/3:

b = a/2

elif a <255/3*2:

b = (a-255/3)*2 + 255/6

else:

b = (a-255/3*2)/2 + 255/6 +255/3*2

return b

rows = img.shape[0]

cols = img.shape[1]

cover = copy.deepcopy(gray)

for i in range(rows):

for j in range(cols):

cover[i][j] = chng(cover[i][j])

CV2.imwrite('cover.png', cover)

下一步是直方图均衡化

# histogram equalization

def hist_equal(img, z_max=255):

H, W = img.shape

# S is the total of pixels

S = H * W * 1.

out = img.copy()

sum_h = 0.

for i in range(1, 255):

ind = np.where(img == i)

sum_h += len(img[ind])

z_prime = z_max / S * sum_h

out[ind] = z_prime

out = out.astype(np.uint8)

return out

covereq = hist_equal(cover)

CV2.imwrite('covereq.png', covereq)

在实现滤波之前先添加高斯噪声和椒盐噪声（代码来源于网络）

不知道这个椒盐噪声的名字是谁起的感觉隔壁小孩都馋哭了

用到了random.gauss()

percentage是噪声占比

def GaussianNoise(src,means,sigma,percetage):

NoiseImg=src

NoiseNum=int(percetage*src.shape[0]*src.shape[1])

for i in range(NoiseNum):

randX=random.randint(0,src.shape[0]-1)

randY=random.randint(0,src.shape[1]-1)

NoiseImg[randX, randY]=NoiseImg[randX,randY]+random.gauss(means,sigma)

if NoiseImg[randX, randY]<0:

NoiseImg[randX, randY]=0

elif NoiseImg[randX, randY]>255:

NoiseImg[randX, randY]=255

return NoiseImg

def PepperandSalt(src,percetage):

NoiseImg=src

NoiseNum=int(percetage*src.shape[0]*src.shape[1])

for i in range(NoiseNum):

randX=random.randint(0,src.shape[0]-1)

randY=random.randint(0,src.shape[1]-1)

if random.randint(0,1)<=0.5:

NoiseImg[randX,randY]=0

else:

NoiseImg[randX,randY]=255

return NoiseImg

covereqg = GaussianNoise(covereq, 2, 4, 0.8)

CV2.imwrite('covereqg.png', covereqg)

covereqps = PepperandSalt(covereq, 0.05)

CV2.imwrite('covereqps.png', covereqps)

下面开始均值滤波和中值滤波了

就以n x n为例，均值滤波就是用这n x n个像素点灰度值的平均值代替中心点，而中值就是中位数代替中心点，边界点周围补0；前两个函数的作用是算出这个点的灰度值，后两个是对整张图片进行

#均值滤波模板

def mean_filter(x, y, step, img):

sum_s = 0

for k in range(x-int(step/2), x+int(step/2)+1):

for m in range(y-int(step/2), y+int(step/2)+1):

if k-int(step/2) 0 or k+int(step/2)+1 >img.shape[0]

or m-int(step/2) 0 or m+int(step/2)+1 >img.shape[1]:

sum_s += 0

else:

sum_s += img[k][m] / (step*step)

return sum_s

#中值滤波模板

def median_filter(x, y, step, img):

sum_s=[]

for k in range(x-int(step/2), x+int(step/2)+1):

for m in range(y-int(step/2), y+int(step/2)+1):

if k-int(step/2) 0 or k+int(step/2)+1 >img.shape[0]

or m-int(step/2) 0 or m+int(step/2)+1 >img.shape[1]:

sum_s.append(0)

else:

sum_s.append(img[k][m])

sum_s.sort()

return sum_s[(int(step*step/2)+1)]

def median_filter_go(img, n):

img1 = copy.deepcopy(img)

for i in range(img.shape[0]):

for j in range(img.shape[1]):

img1[i][j] = median_filter(i, j, n, img)

return img1

def mean_filter_go(img, n):

img1 = copy.deepcopy(img)

for i in range(img.shape[0]):

for j in range(img.shape[1]):

img1[i][j] = mean_filter(i, j, n, img)

return img1

完整main代码如下：

if __name__ == "__main__":

# 读入原始图像

img = CV2.imread('joi.jpg')

# 灰度化处理

gray = CV2.cvtColor(img, CV2.COLOR_BGR2GRAY)

CV2.imwrite('img.png', gray)

rows = img.shape[0]

cols = img.shape[1]

cover = copy.deepcopy(gray)

for i in range(rows):

for j in range(cols):

cover[i][j] = chng(cover[i][j])

CV2.imwrite('cover.png', cover)

covereq = hist_equal(cover)

CV2.imwrite('covereq.png', covereq)

covereqg = GaussianNoise(covereq, 2, 4, 0.8)

CV2.imwrite('covereqg.png', covereqg)

covereqps = PepperandSalt(covereq, 0.05)

CV2.imwrite('covereqps.png', covereqps)

meanimg3 = mean_filter_go(covereqps, 3)

CV2.imwrite('medimg3.png', meanimg3)

meanimg5 = mean_filter_go(covereqps, 5)

CV2.imwrite('meanimg5.png', meanimg5)

meanimg7 = mean_filter_go(covereqps, 7)

CV2.imwrite('meanimg7.png', meanimg7)

medimg3 = median_filter_go(covereqg, 3)

CV2.imwrite('medimg3.png', medimg3)

medimg5 = median_filter_go(covereqg, 5)

CV2.imwrite('medimg5.png', medimg5)

medimg7 = median_filter_go(covereqg, 7)

CV2.imwrite('medimg7.png', medimg7)

medimg4 = median_filter_go(covereqps, 7)

CV2.imwrite('medimg4.png', medimg4)

---