图像的基本操作--反转、放大等

1.打开图像

 

import os
import matplotlib.image as mpimg
import numpy as np
import matplotlib.pyplot as plt #显示照片
from PIL import Image

cwd = os.getcwd()   # 获得当前工作路径
print(cwd)
path = cwd + "/images"
valid_exts = [".jpg", ".gif", ".png", ".tga", ".jpeg"]

print('%d files in %s'%(len(os.listdir(path)), path))

imgs =[]
names =[]
for f in os.listdir(path):
    ext = os.path.splitext(f)[1]   #ext:('1', '.jpg')  ext[1]:  .jpg
    if ext.lower() not in valid_exts :
        continue
    fullpath = os.path.join(path, f)
    imgs.append(mpimg.imread(fullpath))
    names.append(os.path.splitext(f)[0]+os.path.splitext(f)[1])


print('%d images loaded'%(len(imgs)))

nimgs = len (imgs)
randidx = np.sort(np.random.randint(nimgs, size=3))
print(randidx)

print('Type of imgs :', type(imgs))
print('Length of imgs :', len(imgs))

for curr_img, curr_name, i in zip([imgs[j] for j in randidx], [names[j] for j in randidx], range(len(randidx))):
    print('[%d] Type of curr_img: %s'%(i, type(curr_img)))
    print('Name is :%s'%(curr_name))
    print('size of curr_img: %s '%(curr_img.shape,)) # ,逗号

import matplotlib.image as mpimg

mpimg.imread()

可以用 im.shape

from PIL import Image

利用Image

 

valid_exts1=["jpg", "gif", "png", "tga", "jpeg"]
ims = input('请输入文件名：')
print(ims.split('.', 1)[1])
while ims.split('.', 1)[1] in valid_exts1:
    im = mpimg.imread(os.path.join(path, ims ))
    plt.imshow(im)
    print(im.shape)
    plt.show()
    break

2. 旋转（PIL）

#旋转
im = Image.open(os.path.join(path, ims ))
angle = int(input('请输入旋转角度：'))
im_rotate = im.rotate(angle=angle)
im_rotate.show()
plt.imshow(im_rotate)
plt.show()

3.缩放

p = float(input('请输入缩放比例（0-1）:'))
(x, y) = im.size
x_s = x *p
y_s = y *p
out = im.resize((x_s, y_s), Image.ANTIALIAS)
out.show()
#im_resize = im.resize((128, 128)) 直接输入所需的大小

补充：os.walk()用类似于深度遍历的方式遍历文件夹中的子文件夹以及文件。最基本的显示方式为(root_path,[file_dirs],[files])，以元组为单位区分每一层的，每一层又分成三个部分根目录路径、本目录中文件夹名字、本目录中文件名字。

for root, dirs, files in os.walk(image_path):

        print("root:"+root)

        print(dirs)

        print(files)

4.水平反转

out1 = im.transpose(Image.FLIP_LEFT_RIGHT)  # 实现翻转
out1.show()

 5.上下反转

out2 = im.transpose(Image.FLIP_TOP_BOTTOM)  #上下反转

out2.show()

 

 6.提取轮廓
cv2.imread()-----opencv读取图像为b,g,r方法

灰度化与二值化的区别

灰度化：在RGB模型中，如果R=G=B时，则彩色表示一种灰度颜色，其中R=G=B的值叫灰度值，因此，灰度图像每个像素只需一个字节存放灰度值（又称强度值、亮度值），灰度范围为0-255。一般常用的是加权平均法来获取每个像素点的灰度值。

二值化：图像的二值化，就是将图像上的像素点的灰度值设置为0或255，也就是将整个图像呈现出明显的只有黑和白的视觉效果

img = cv2.imread(r'F:\pytorch Tutorial\study\images\1.jpg')
cv2.imshow('Example',img)
cv2.waitKey(0)

gray_img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) #灰度化
cv2.imshow('Example1',gray_img)
cv2.waitKey(0)

#二值化
def get_binary_img(img):
    # gray img to bin image
    bin_img = np.zeros(shape=(img.shape), dtype=np.uint8)
    h = img.shape[0]
    w = img.shape[1]
    for i in range(h):
        for j in range(w):
            bin_img[i][j] = 255 if img[i][j] > 127 else 0
    return bin_img
# 调用
bin_img = get_binary_img(gray_img)

cv2.imshow('Example2',bin_img )
cv2.waitKey(0)


def get_contour(bin_img):
    # get contour
    contour_img = np.zeros(shape=(bin_img.shape),dtype=np.uint8)
    contour_img += 255
    h = bin_img.shape[0]
    w = bin_img.shape[1]
    for i in range(1,h-1):
        for j in range(1,w-1):
            if(bin_img[i][j]==0):
                contour_img[i][j] = 0
                sum = 0
                sum += bin_img[i - 1][j + 1] #判断此像素点附近八个点的值是否为0
                sum += bin_img[i][j + 1]
                sum += bin_img[i + 1][j + 1]
                sum += bin_img[i - 1][j]
                sum += bin_img[i + 1][j]
                sum += bin_img[i - 1][j - 1]
                sum += bin_img[i][j - 1]
                sum += bin_img[i + 1][j - 1]
                if sum ==  0:
                    contour_img[i][j] = 255

    return contour_img
# 调用
contour_img = get_contour(bin_img)

cv2.imshow('Example3',contour_img)
cv2.waitKey(0)

7.灰度图反相处理

#反相处理
#(x_1,y_1)=gray_img.size

img = cv2.imread(r'F:\pytorch Tutorial\study\images\1.jpg')
cv2.imshow('Example',img)
cv2.waitKey(0)

gray_img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) #灰度化
cv2.imshow('Example1',gray_img)
cv2.waitKey(0)
a=np.ones_like(gray_img, dtype=np.int32) *255
b = a- gray_img 
plt.imshow(b,cmap='gray')
plt.show()

 8.python直方图均衡化

直方图均衡化就是把一已知灰度概率分布的图像经过一种变换使之演变成一幅具有均匀灰度概率分布的新图像

灰度处理

问题1：如何计算直方图？

对于8位灰度图来说，颜色有256级，统计每级的个数，然后把结果用图表表示出来就可以了

 

import cv2
import matplotlib.pyplot as plt
import numpy as np
import os


gray_level = 256

def piexl_im(img):

    assert isinstance(img, np.ndarray)
    (height, width) = img.shape
    prob = np.zeros(shape=(256))

    for i in range(height):
        for j in range(width):
            k = img[i,j]   # 一点的像素值

            prob[k] = prob[k]+1
    return prob


def piexl_prob(img):

    assert isinstance(img, np.ndarray)
    (height, width) = img.shape
    prob = np.zeros(shape=(256))

    for i in range(height):
        for j in range(width):
            k = img[i,j]   # 一点的像素值

            prob[k] = prob[k]+1

    prob= prob/(height * width)
    return prob


img = cv2.imread('1.jpg')
gray_img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY) #灰度化

x=piexl_im(gray_img)

a= range(256)
plt.bar(a, x,width=0.2)
plt.show()

probably = piexl_prob(gray_img)

def probability_to_histogram(img, prob):
    prob = np.cumsum(prob)  # 累计概率

    img_map = [int(255 * prob[i] + 0.5) for i in range(256)]  # 像素值映射

    # 像素值替换
    assert isinstance(img, np.ndarray)
    r, c = img.shape
    for ri in range(r):
        for ci in range(c):
            img[ri, ci] = img_map[img[ri, ci]]

    return img

def plot(y, name):
    """
    画直方图，len(y)==gray_level
    :param y: 概率值
    :param name:
    :return:
    """
    plt.figure(num=name)
    plt.bar([i for i in range(gray_level)], y, width=1)


img = cv2.imread("1.jpg", 0)  # 读取灰度图

prob = piexl_prob(img)
plot(prob, "原图直方图")

# 直方图均衡化
img = probability_to_histogram(img, prob)
cv2.imwrite("source_hist.jpg", img)  # 保存图像

prob = piexl_prob(img)
plot(prob, "直方图均衡化结果")

plt.show()

补充： cumsum（）

9.均值滤波、高斯滤波

 cv.blur( 图像， （核3，3））------》改变核的大小可以改变滤波强度

高斯滤波

#自定义均值滤波

def mean_blur(img,kernel_size=3):
    h, w, c = img.shape

    #零填充
    k = kernel_size//2
    a = np.zeros((h + 2*k,w + 2*k,c),dtype=np.float)
    a[k:k + h, k:k + w] = img.copy().astype(np.float)

    #卷积
    temp= a.copy()
    for y in range(h):
        for x in range(w):
            for ci in range(c):
                a[k+y,k+x,ci]=np.mean(temp[y:y+kernel_size,x:x+kernel_size,ci])

    out = a[k:k + h, k:k + w].astype(np.uint8)
    return out

#自定义高斯滤波
def gaussian_filter(img, K_size=3, sigma=1.3):
    if len(img.shape) == 3:

        H, W, C = img.shape

    else:

        img = np.expand_dims(img, axis=-1)

        H, W, C = img.shape

    ## Zero padding

    pad = K_size // 2

    out = np.zeros((H + pad * 2, W + pad * 2, C), dtype=np.float)

    out[pad: pad + H, pad: pad + W] = img.copy().astype(np.float)

    ## prepare Kernel

    K = np.zeros((K_size, K_size), dtype=np.float)

    for x in range(-pad, -pad + K_size):

        for y in range(-pad, -pad + K_size):
            K[y + pad, x + pad] = np.exp(-(x ** 2 + y ** 2) / (2 * (sigma ** 2)))

    K /= (2 * np.pi * sigma * sigma)

    K /= K.sum()

    tmp = out.copy()

    # filtering

    for y in range(H):

        for x in range(W):

            for c in range(C):
                out[pad + y, pad + x, c] = np.sum(K * tmp[y: y + K_size, x: x + K_size, c])

    out = np.clip(out, 0, 255)

    out = out[pad: pad + H, pad: pad + W].astype(np.uint8)

    return out

参考文献：

Python中读取图片的6种方式_Python_脚本之家

 实例说明图像的灰度化和二值化的区别 - 云+社区 - 腾讯云 

【深度好文】Python图像处理之目标物体轮廓提取_sgzqc的专栏-CSDN博客                  Opencv-python（cv2）图像读取、显示与保存，看这一篇就够了_风雪夜归人o的博客-CSDN博客 

数字图像处理(11): 图像平滑 (均值滤波、中值滤波和高斯滤波)_TechArtisan6的博客-CSDN博客

手动实现均值滤波（python）_陨星落云的博客-CSDN博客_均值滤波python

高斯滤波详解 python实现高斯滤波_Ibelievesunshine的博客-CSDN博客

(29条消息) 【图像处理算法】直方图均衡化_GQ-CSDN博客_直方图均衡化

(29条消息) 直方图均衡化-python实现_lsh呵呵-CSDN博客_python直方图均衡化