好用！Numpy处理图像的几个小技巧

来自：Deephub Imba

图像处理是一种数学计算。数字图像由称为像素的彩色小点组成。每个像素由红、绿、蓝(RGB)三个独立的颜色组成。每个像素中的主色由每个RGB分量的数值决定。

本文将介绍10个使用使用NumPy就可以进行的图像处理步骤，虽然有更强大的图像处理库，但是这些简单的方法可以让我们更加熟练的掌握NumPy的操作。

我们首先使用pillow读取图像

 import numpy as np
 
 #Use PIL to access image data
 from PIL import Image
 img = Image.open('monalisa.jpg')
 
 #Create array from image data 
 M = np.array(img)
 
 #Display array from image data 
 display(Image.fromarray(M))

1、缩小图像

 def reduce_image_size_by_n(image, n):
 
     # Get the height and width of the image
     height, width, channels = image.shape
 
     # Reduce the height and width by n
     new_height = height // n
     new_width = width // n
 
     # Create a new array to store the reduced image
     downsampled_image = np.zeros((new_height, new_width, channels), dtype=image.dtype)
 
     # Iterate over each pixel of the reduced image
     for i in range(new_height):
         for j in range(new_width):
 
             # Take every other pixel along each axis to reduce the image
 
             downsampled_image[i, j] = image[n*i, n*j]
 
     return downsampled_image
 
 #Try the function using n = 2
 
 reduced_M = reduce_image_size_by_n(M, 2)
 
 display(reduced_M)

2、水平翻转

 def flip_image(image):
 
   # Takes all rows in image (:) and reverses it the order of columns (::-1)
   flip_image = image[:, ::-1]
   return flip_image
 
 #Try function using reduced image
 display(flip_image(reduced_M))

3、垂直翻转

 def rotate_image (image, n):
   # rotate image using rot90, use n to determine number of rotation 
   rotated_img = Image.fromarray(np.rot90(image, k=n, axes=(1, 0)))
   return rotated_img
 
 #rotate image twice (n=2)
 display(rotate_image(reduced_M, 2))

4、裁剪图像

 def crop_image(image, crop_ratio, zoom_ratio):
 
   #create focused part using crop_ratio and zoom_ratio of choice
   
   top = image.shape[0] // crop_ratio 
   bottom = zoom_ratio * image.shape[0] // crop_ratio
   left = image.shape[1] // crop_ratio
   right = zoom_ratio * image.shape[1] // crop_ratio
 
   # Extract the focused part using array slicing
   focused_part = image[top:bottom, left:right]
   return focused_part
 
 display(crop_image(reduced_M, 4, 2))

5、RGB通道

 def RGB_image(image,image_color):
 
   if image_color == 'R':
     #make a copy of image for the color channel
     img_R = image.copy()
     #set other color channel to zero. Here Red is the first channel [0] 
     img_R[:, :, (1, 2)] = 0
     return img_R
 
   elif image_color == 'G':
     img_G = image.copy()
     #set other color channel to zero. Here Green is the second channel [1]
     img_G[:, :, (0, 2)] = 0
     return img_G
 
   elif image_color == 'B':
     img_B = image.copy()
     #set other color channel to zero. Here Blue is the third channel [2]
     img_B[:, :, (0, 1)] = 0
     return img_B

查看红色通道

 M_red = Image.fromarray(RGB_image(reduced_M, 'R'))
 
 display(M_red)

绿色

 M_green = Image.fromarray(RGB_image(reduced_M, 'G'))
 
 display(M_green)

蓝色

 M_blue = Image.fromarray(RGB_image(reduced_M, 'B'))
 
 display(M_blue)

6、应用滤镜

这里使用棕褐色(Sepia)作为示例，可以根据不同的要求修改转换矩阵

 def apply_sepia(image):
     # Sepia transformation matrix
     sepia_matrix = np.array([[0.393, 0.769, 0.189],
                              [0.349, 0.686, 0.168],
                              [0.272, 0.534, 0.131]])
 
     # Apply the sepia transformation
     sepia_img = image.dot(sepia_matrix.T)  # Using matrix multiplication
 
     # Ensure values are within valid range [0, 255]
     sepia_img = np.clip(sepia_img, 0, 255)
 
     return sepia_img.astype(np.uint8)
 
 # Apply sepia effect
 M_sepia = Image.fromarray(apply_sepia(reduced_M))
 
 display(M_sepia)

7、灰度化

灰度化可以简单的理解为将RBG三个通道合并成一个黑白的通道

 import numpy as np
 
 def grayscale(image):
     # Convert the RGB image to grayscale using weighted average
     grayscale_img = np.dot(image[..., :3], [0.2989, 0.5870, 0.1140])
 
     # Ensure values are within valid range [0, 255]
     grayscale_img = np.clip(grayscale_img, 0, 255)
 
     # Convert to uint8 data type
     grayscale_img = grayscale_img.astype(np.uint8)
 
     return grayscale_img
 
 # Convert the image to grayscale
 M_gray = grayscale(reduced_M)
 
 display(M_gray)

8、像素化

像素是一个一个色块组成的，像素化顾名思义就是将图像分成一定的区域，并将这些区域转换成相应的色块，再有色块构成图形。类似于色彩构图。简单来说,就是把矢量图形转换成像素点组成的点阵图形，也叫栅格化。

 def pixelate_image(image, block_size):
   
     # Determine the number of blocks in each dimension
     num_blocks_y = image.shape[0] // block_size
     num_blocks_x = image.shape[1] // block_size
 
     # Calculate the average color for each block
     block_means = np.zeros((num_blocks_y, num_blocks_x, 3), dtype=np.uint8)
     for y in range(num_blocks_y):
         for x in range(num_blocks_x):
             block = image[y * block_size: (y + 1) * block_size,
                         x * block_size: (x + 1) * block_size]
             block_mean = np.mean(block, axis=(0, 1))
             block_means[y, x] = block_mean.astype(np.uint8)
 
     # Upsample block means to original image size
     pixelated_image = np.repeat(np.repeat(block_means, block_size, axis=0), block_size, axis=1)
 
     return pixelated_image
 
 
 # Set the block size for pixelation (adjust as needed)
 block_size = 10
 
 # Pixelate the image
 M_pixelated = Image.fromarray(pixelate_image(reduced_M, block_size))
 
 display(M_pixelated)

更通俗的的讲就是我的世界风格的图像

9、二值化（Binarize）

二值化是将数值型特征取值阈值化转换为布尔型特征取值，或者通俗的讲就是设定一个阈值，超过阈值设置成ture，否则设置成false

 def binarize_image(image, threshold):
 
   #set pixel value greater than threshold to 255
   binarize_image = ((image > threshold) * 255).astype(np.uint8)
   
   return binarize_image
 
 #set threshold
 threshold = 68
 
 M_binarized = Image.fromarray(binarize_image(reduced_M, threshold))
 
 display(M_binarized)

10、图像融合

最简单的图像同和方法就是根据不同的透明度，对2张图象的像素求和相加，如下所示

 #import and resize second image 
 
 img_2 = np.array(Image.open('Eiffel.jpg').resize(reduced_M.shape[1::-1]))  
 
 def blend_image(image1, image2, , visibility_2 ):
 
   #blend images by multiplying by visibility ratio for each image
 
   blend_image = (image1 * visibility_1 + image2 * visibility_2).astype(np.uint8)
 
   return blend_image
 
 
 
 modified_image = Image.fromarray(blend_image(reduced_M, img_2, 0.7, 0.3))
 
 display(modified_image)

总结

对于图像的操作其实就是对于图像进行数组操作的过程，我们这里展示的一些简单的操作只是为了熟悉Numpy的操作，如果需要更加专业的操作请使用更加专业的库，例如OpenCV或者Pillow。

作者：Ayo Akinkugbe

往期文章

推荐3款自动爬虫神器，再也不用手撸代码了
新式爬虫利器，网页解锁能力非常强大！
使用八爪鱼爬虫+Kimi AI分析小米SU7舆情数据，终于知道它为什么火了
推荐我常用的爬虫工具，三种爬虫方式，搞定反爬和动态页面6个强大且流行的Python爬虫库，强烈推荐！