{"id":1115807,"date":"2025-01-08T18:10:19","date_gmt":"2025-01-08T10:10:19","guid":{"rendered":"https:\/\/docs.pingcode.com\/ask\/ask-ask\/1115807.html"},"modified":"2025-01-08T18:10:23","modified_gmt":"2025-01-08T10:10:23","slug":"%e5%a6%82%e4%bd%95%e7%94%a8python%e7%bb%99%e7%81%b0%e5%ba%a6%e5%9b%be%e5%83%8f%e5%8e%bb%e5%99%aa","status":"publish","type":"post","link":"https:\/\/docs.pingcode.com\/ask\/1115807.html","title":{"rendered":"\u5982\u4f55\u7528PYTHON\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a"},"content":{"rendered":"

\"\u5982\u4f55\u7528PYTHON\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a\"<\/p>\n

\u5982\u4f55\u7528Python\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a<\/strong><\/p>\n<\/p>\n

\u4f7f\u7528Python\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a\u7684\u65b9\u6cd5\u5305\u62ec\u5747\u503c\u6ee4\u6ce2\u3001\u4e2d\u503c\u6ee4\u6ce2\u3001\u9ad8\u65af\u6ee4\u6ce2\u3001\u53cc\u8fb9\u6ee4\u6ce2<\/strong>\u7b49\u3002\u8fd9\u4e9b\u65b9\u6cd5\u5404\u6709\u4f18\u52a3\uff0c\u5176\u4e2d\u9ad8\u65af\u6ee4\u6ce2\u88ab\u5e7f\u6cdb\u5e94\u7528\u3002\u9ad8\u65af\u6ee4\u6ce2\u5229\u7528\u6b63\u6001\u5206\u5e03\u51fd\u6570\u5bf9\u56fe\u50cf\u8fdb\u884c\u5e73\u6ed1\u5904\u7406\uff0c\u6709\u6548\u53bb\u9664\u566a\u58f0\u7684\u540c\u65f6\u4fdd\u7559\u56fe\u50cf\u7684\u8fb9\u7f18\u7ec6\u8282\u3002<\/p>\n<\/p>\n

\u9ad8\u65af\u6ee4\u6ce2<\/strong>\u901a\u8fc7\u5728\u56fe\u50cf\u4e0a\u5e94\u7528\u9ad8\u65af\u51fd\u6570\u6765\u5e73\u6ed1\u56fe\u50cf\uff0c\u6709\u6548\u51cf\u5c11\u566a\u58f0\u3002\u9ad8\u65af\u6ee4\u6ce2\u5668\u7684\u53c2\u6570\u5305\u62ec\u6ee4\u6ce2\u5668\u7684\u5c3a\u5bf8\u548c\u6807\u51c6\u5dee\uff0c\u5c3a\u5bf8\u51b3\u5b9a\u6ee4\u6ce2\u5668\u7684\u8986\u76d6\u8303\u56f4\uff0c\u6807\u51c6\u5dee\u51b3\u5b9a\u6ee4\u6ce2\u7684\u5f3a\u5ea6\u3002\u9009\u53d6\u5408\u9002\u7684\u53c2\u6570\uff0c\u53ef\u4ee5\u5728\u53bb\u566a\u548c\u4fdd\u6301\u56fe\u50cf\u7ec6\u8282\u4e4b\u95f4\u53d6\u5f97\u5e73\u8861\u3002<\/p>\n<\/p>\n

\u4e00\u3001\u5747\u503c\u6ee4\u6ce2<\/h3>\n<\/p>\n

\u5747\u503c\u6ee4\u6ce2\u662f\u4e00\u79cd\u7b80\u5355\u4e14\u5e38\u7528\u7684\u56fe\u50cf\u5e73\u6ed1\u65b9\u6cd5\u3002\u5b83\u901a\u8fc7\u5c06\u56fe\u50cf\u4e2d\u7684\u6bcf\u4e2a\u50cf\u7d20\u66ff\u6362\u4e3a\u5176\u90bb\u57df\u5185\u50cf\u7d20\u7684\u5747\u503c\u6765\u5b9e\u73b0\u53bb\u566a\u3002\u5747\u503c\u6ee4\u6ce2\u7684\u4f18\u70b9\u662f\u8ba1\u7b97\u7b80\u5355\uff0c\u4f46\u7f3a\u70b9\u662f\u5bb9\u6613\u6a21\u7cca\u56fe\u50cf\u7ec6\u8282\uff0c\u7279\u522b\u662f\u8fb9\u7f18\u90e8\u5206\u3002<\/p>\n<\/p>\n

\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n

\u4f7f\u7528OpenCV\u5e93\u53ef\u4ee5\u65b9\u4fbf\u5730\u5b9e\u73b0\u5747\u503c\u6ee4\u6ce2\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff1a<\/p>\n<\/p>\n

import cv2<\/p>\n
import numpy as np<\/p>\n
\u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf<\/strong><\/h2>\n
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)<\/p>\n
\u5e94\u7528\u5747\u503c\u6ee4\u6ce2<\/strong><\/h2>\n
kernel_size = 5  # \u6ee4\u6ce2\u5668\u5c3a\u5bf8<\/p>\n
blurred_image = cv2.blur(image, (kernel_size, kernel_size))<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
cv2.imshow('Original', image)<\/p>\n
cv2.imshow('Blurred', blurred_image)<\/p>\n
cv2.wAI<\/a>tKey(0)<\/p>\n
cv2.destroyAllWindows()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u4e86cv2.blur<\/code>\u51fd\u6570\u5bf9\u56fe\u50cf\u8fdb\u884c\u5747\u503c\u6ee4\u6ce2\u3002kernel_size<\/code>\u51b3\u5b9a\u4e86\u6ee4\u6ce2\u5668\u7684\u5c3a\u5bf8\uff0c\u53ef\u4ee5\u6839\u636e\u9700\u8981\u8c03\u6574\u3002<\/p>\n<\/p>\n
\u4e8c\u3001\u4e2d\u503c\u6ee4\u6ce2<\/h3>\n<\/p>\n
\u4e2d\u503c\u6ee4\u6ce2\u662f\u4e00\u79cd\u975e\u7ebf\u6027\u6ee4\u6ce2\u65b9\u6cd5\uff0c\u5e38\u7528\u4e8e\u53bb\u9664\u6912\u76d0\u566a\u58f0\u3002\u5b83\u901a\u8fc7\u5c06\u56fe\u50cf\u4e2d\u7684\u6bcf\u4e2a\u50cf\u7d20\u66ff\u6362\u4e3a\u5176\u90bb\u57df\u5185\u50cf\u7d20\u7684\u4e2d\u503c\u6765\u5b9e\u73b0\u53bb\u566a\u3002\u4e2d\u503c\u6ee4\u6ce2\u7684\u4f18\u70b9\u662f\u80fd\u6709\u6548\u53bb\u9664\u5b64\u7acb\u7684\u566a\u58f0\u70b9\uff0c\u540c\u65f6\u4fdd\u7559\u56fe\u50cf\u7684\u8fb9\u7f18\u7ec6\u8282\u3002<\/p>\n<\/p>\n
\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n
\u4f7f\u7528OpenCV\u5e93\u53ef\u4ee5\u65b9\u4fbf\u5730\u5b9e\u73b0\u4e2d\u503c\u6ee4\u6ce2\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff1a<\/p>\n<\/p>\n
# \u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf<\/p>\n
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)<\/p>\n
\u5e94\u7528\u4e2d\u503c\u6ee4\u6ce2<\/strong><\/h2>\n
kernel_size = 5  # \u6ee4\u6ce2\u5668\u5c3a\u5bf8<\/p>\n
median_blurred_image = cv2.medianBlur(image, kernel_size)<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
cv2.imshow('Original', image)<\/p>\n
cv2.imshow('Median Blurred', median_blurred_image)<\/p>\n
cv2.waitKey(0)<\/p>\n
cv2.destroyAllWindows()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u4e86cv2.medianBlur<\/code>\u51fd\u6570\u5bf9\u56fe\u50cf\u8fdb\u884c\u4e2d\u503c\u6ee4\u6ce2\u3002kernel_size<\/code>\u51b3\u5b9a\u4e86\u6ee4\u6ce2\u5668\u7684\u5c3a\u5bf8\uff0c\u53ef\u4ee5\u6839\u636e\u9700\u8981\u8c03\u6574\u3002<\/p>\n<\/p>\n
\u4e09\u3001\u9ad8\u65af\u6ee4\u6ce2<\/h3>\n<\/p>\n
\u9ad8\u65af\u6ee4\u6ce2\u662f\u4e00\u79cd\u7ebf\u6027\u6ee4\u6ce2\u65b9\u6cd5\uff0c\u901a\u8fc7\u5728\u56fe\u50cf\u4e0a\u5e94\u7528\u9ad8\u65af\u51fd\u6570\u6765\u5e73\u6ed1\u56fe\u50cf\uff0c\u6709\u6548\u51cf\u5c11\u566a\u58f0\u3002\u9ad8\u65af\u6ee4\u6ce2\u5668\u7684\u53c2\u6570\u5305\u62ec\u6ee4\u6ce2\u5668\u7684\u5c3a\u5bf8\u548c\u6807\u51c6\u5dee\uff0c\u5c3a\u5bf8\u51b3\u5b9a\u6ee4\u6ce2\u5668\u7684\u8986\u76d6\u8303\u56f4\uff0c\u6807\u51c6\u5dee\u51b3\u5b9a\u6ee4\u6ce2\u7684\u5f3a\u5ea6\u3002\u9009\u53d6\u5408\u9002\u7684\u53c2\u6570\uff0c\u53ef\u4ee5\u5728\u53bb\u566a\u548c\u4fdd\u6301\u56fe\u50cf\u7ec6\u8282\u4e4b\u95f4\u53d6\u5f97\u5e73\u8861\u3002<\/p>\n<\/p>\n
\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n
\u4f7f\u7528OpenCV\u5e93\u53ef\u4ee5\u65b9\u4fbf\u5730\u5b9e\u73b0\u9ad8\u65af\u6ee4\u6ce2\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff1a<\/p>\n<\/p>\n
# \u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf<\/p>\n
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)<\/p>\n
\u5e94\u7528\u9ad8\u65af\u6ee4\u6ce2<\/strong><\/h2>\n
kernel_size = (5, 5)  # \u6ee4\u6ce2\u5668\u5c3a\u5bf8<\/p>\n
sigma = 1.5  # \u6807\u51c6\u5dee<\/p>\n
gaussian_blurred_image = cv2.GaussianBlur(image, kernel_size, sigma)<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
cv2.imshow('Original', image)<\/p>\n
cv2.imshow('Gaussian Blurred', gaussian_blurred_image)<\/p>\n
cv2.waitKey(0)<\/p>\n
cv2.destroyAllWindows()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u4e86cv2.GaussianBlur<\/code>\u51fd\u6570\u5bf9\u56fe\u50cf\u8fdb\u884c\u9ad8\u65af\u6ee4\u6ce2\u3002kernel_size<\/code>\u51b3\u5b9a\u4e86\u6ee4\u6ce2\u5668\u7684\u5c3a\u5bf8\uff0csigma<\/code>\u51b3\u5b9a\u4e86\u9ad8\u65af\u5206\u5e03\u7684\u6807\u51c6\u5dee\uff0c\u53ef\u4ee5\u6839\u636e\u9700\u8981\u8c03\u6574\u3002<\/p>\n<\/p>\n
\u56db\u3001\u53cc\u8fb9\u6ee4\u6ce2<\/h3>\n<\/p>\n
\u53cc\u8fb9\u6ee4\u6ce2\u662f\u4e00\u79cd\u975e\u7ebf\u6027\u6ee4\u6ce2\u65b9\u6cd5\uff0c\u901a\u8fc7\u540c\u65f6\u8003\u8651\u7a7a\u95f4\u8ddd\u79bb\u548c\u50cf\u7d20\u503c\u5dee\u5f02\u6765\u5e73\u6ed1\u56fe\u50cf\uff0c\u6709\u6548\u51cf\u5c11\u566a\u58f0\u3002\u4e0e\u5176\u4ed6\u6ee4\u6ce2\u65b9\u6cd5\u76f8\u6bd4\uff0c\u53cc\u8fb9\u6ee4\u6ce2\u80fd\u591f\u66f4\u597d\u5730\u4fdd\u7559\u56fe\u50cf\u7684\u8fb9\u7f18\u7ec6\u8282\u3002<\/p>\n<\/p>\n
\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n
\u4f7f\u7528OpenCV\u5e93\u53ef\u4ee5\u65b9\u4fbf\u5730\u5b9e\u73b0\u53cc\u8fb9\u6ee4\u6ce2\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff1a<\/p>\n<\/p>\n
# \u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf<\/p>\n
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)<\/p>\n
\u5e94\u7528\u53cc\u8fb9\u6ee4\u6ce2<\/strong><\/h2>\n
diameter = 9  # \u6ee4\u6ce2\u5668\u76f4\u5f84<\/p>\n
sigma_color = 75  # \u989c\u8272\u7a7a\u95f4\u7684\u6807\u51c6\u5dee<\/p>\n
sigma_space = 75  # \u5750\u6807\u7a7a\u95f4\u7684\u6807\u51c6\u5dee<\/p>\n
bilateral_blurred_image = cv2.bilateralFilter(image, diameter, sigma_color, sigma_space)<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
cv2.imshow('Original', image)<\/p>\n
cv2.imshow('Bilateral Blurred', bilateral_blurred_image)<\/p>\n
cv2.waitKey(0)<\/p>\n
cv2.destroyAllWindows()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u4e86cv2.bilateralFilter<\/code>\u51fd\u6570\u5bf9\u56fe\u50cf\u8fdb\u884c\u53cc\u8fb9\u6ee4\u6ce2\u3002diameter<\/code>\u51b3\u5b9a\u4e86\u6ee4\u6ce2\u5668\u7684\u76f4\u5f84\uff0csigma_color<\/code>\u548csigma_space<\/code>\u51b3\u5b9a\u4e86\u989c\u8272\u7a7a\u95f4\u548c\u5750\u6807\u7a7a\u95f4\u7684\u6807\u51c6\u5dee\uff0c\u53ef\u4ee5\u6839\u636e\u9700\u8981\u8c03\u6574\u3002<\/p>\n<\/p>\n
\u4e94\u3001\u9891\u57df\u6ee4\u6ce2<\/h3>\n<\/p>\n
\u9891\u57df\u6ee4\u6ce2\u662f\u4e00\u79cd\u901a\u8fc7\u5085\u91cc\u53f6\u53d8\u6362\u5c06\u56fe\u50cf\u4ece\u7a7a\u95f4\u57df\u8f6c\u6362\u5230\u9891\u57df\uff0c\u7136\u540e\u5728\u9891\u57df\u4e2d\u8fdb\u884c\u5904\u7406\u7684\u65b9\u6cd5\u3002\u9891\u57df\u6ee4\u6ce2\u53ef\u4ee5\u6709\u6548\u53bb\u9664\u5468\u671f\u6027\u566a\u58f0\uff0c\u5e76\u4fdd\u7559\u56fe\u50cf\u7684\u7ec6\u8282\u3002<\/p>\n<\/p>\n
\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n
\u4f7f\u7528NumPy\u548cOpenCV\u5e93\u53ef\u4ee5\u65b9\u4fbf\u5730\u5b9e\u73b0\u9891\u57df\u6ee4\u6ce2\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff1a<\/p>\n<\/p>\n
import cv2<\/p>\n
import numpy as np<\/p>\n
from matplotlib import pyplot as plt<\/p>\n
\u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf<\/strong><\/h2>\n
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)<\/p>\n
\u5c06\u56fe\u50cf\u8f6c\u6362\u5230\u9891\u57df<\/strong><\/h2>\n
dft = cv2.dft(np.float32(image), flags=cv2.DFT_COMPLEX_OUTPUT)<\/p>\n
dft_shift = np.fft.fftshift(dft)<\/p>\n
\u521b\u5efa\u9891\u57df\u6ee4\u6ce2\u5668<\/strong><\/h2>\n
rows, cols = image.shape<\/p>\n
crow, ccol = rows \/\/ 2, cols \/\/ 2<\/p>\n
mask = np.zeros((rows, cols, 2), np.uint8)<\/p>\n
mask[crow-30:crow+30, ccol-30:ccol+30] = 1<\/p>\n
\u5e94\u7528\u9891\u57df\u6ee4\u6ce2\u5668<\/strong><\/h2>\n
fshift = dft_shift * mask<\/p>\n
f_ishift = np.fft.ifftshift(fshift)<\/p>\n
image_back = cv2.idft(f_ishift)<\/p>\n
image_back = cv2.magnitude(image_back[:, :, 0], image_back[:, :, 1])<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
plt.subplot(121), plt.imshow(image, cmap='gray')<\/p>\n
plt.title('Original Image'), plt.xticks([]), plt.yticks([])<\/p>\n
plt.subplot(122), plt.imshow(image_back, cmap='gray')<\/p>\n
plt.title('Filtered Image'), plt.xticks([]), plt.yticks([])<\/p>\n
plt.show()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u9996\u5148\u5c06\u56fe\u50cf\u8f6c\u6362\u5230\u9891\u57df\uff0c\u7136\u540e\u521b\u5efa\u4e00\u4e2a\u9891\u57df\u6ee4\u6ce2\u5668\uff0c\u5e76\u5e94\u7528\u5230\u9891\u57df\u56fe\u50cf\u4e0a\u3002\u6700\u540e\uff0c\u901a\u8fc7\u9006\u5085\u91cc\u53f6\u53d8\u6362\u5c06\u56fe\u50cf\u8f6c\u6362\u56de\u7a7a\u95f4\u57df\u3002mask<\/code>\u51b3\u5b9a\u4e86\u6ee4\u6ce2\u5668\u7684\u5f62\u72b6\u548c\u5c3a\u5bf8\uff0c\u53ef\u4ee5\u6839\u636e\u9700\u8981\u8c03\u6574\u3002<\/p>\n<\/p>\n
\u516d\u3001\u975e\u5c40\u90e8\u5747\u503c\u6ee4\u6ce2<\/h3>\n<\/p>\n
\u975e\u5c40\u90e8\u5747\u503c\u6ee4\u6ce2\u662f\u4e00\u79cd\u57fa\u4e8e\u56fe\u50cf\u81ea\u76f8\u4f3c\u6027\u7684\u53bb\u566a\u65b9\u6cd5\u3002\u5b83\u901a\u8fc7\u8ba1\u7b97\u56fe\u50cf\u4e2d\u76f8\u4f3c\u5757\u4e4b\u95f4\u7684\u52a0\u6743\u5e73\u5747\u503c\u6765\u53bb\u9664\u566a\u58f0\uff0c\u80fd\u591f\u66f4\u597d\u5730\u4fdd\u7559\u56fe\u50cf\u7684\u7ec6\u8282\u3002<\/p>\n<\/p>\n
\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n
\u4f7f\u7528OpenCV\u5e93\u53ef\u4ee5\u65b9\u4fbf\u5730\u5b9e\u73b0\u975e\u5c40\u90e8\u5747\u503c\u6ee4\u6ce2\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff1a<\/p>\n<\/p>\n
# \u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf<\/p>\n
image = cv2.imread('image.jpg', cv2.IMREAD_GRAYSCALE)<\/p>\n
\u5e94\u7528\u975e\u5c40\u90e8\u5747\u503c\u6ee4\u6ce2<\/strong><\/h2>\n
h = 10  # \u6ee4\u6ce2\u53c2\u6570<\/p>\n
template_window_size = 7  # \u6a21\u677f\u7a97\u53e3\u5c3a\u5bf8<\/p>\n
search_window_size = 21  # \u641c\u7d22\u7a97\u53e3\u5c3a\u5bf8<\/p>\n
nlm_blurred_image = cv2.fastNlMeansDenoising(image, None, h, template_window_size, search_window_size)<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
cv2.imshow('Original', image)<\/p>\n
cv2.imshow('NLM Blurred', nlm_blurred_image)<\/p>\n
cv2.waitKey(0)<\/p>\n
cv2.destroyAllWindows()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u4f7f\u7528\u4e86cv2.fastNlMeansDenoising<\/code>\u51fd\u6570\u5bf9\u56fe\u50cf\u8fdb\u884c\u975e\u5c40\u90e8\u5747\u503c\u6ee4\u6ce2\u3002h<\/code>\u3001template_window_size<\/code>\u548csearch_window_size<\/code>\u51b3\u5b9a\u4e86\u6ee4\u6ce2\u5668\u7684\u53c2\u6570\uff0c\u53ef\u4ee5\u6839\u636e\u9700\u8981\u8c03\u6574\u3002<\/p>\n<\/p>\n
\u4e03\u3001\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u53bb\u566a<\/h3>\n<\/p>\n
\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\uff08CNN\uff09\u662f\u73b0\u4ee3\u56fe\u50cf\u5904\u7406\u4e2d\u7684\u4e00\u79cd\u5f3a\u5927\u5de5\u5177\uff0c\u53ef\u4ee5\u7528\u4e8e\u56fe\u50cf\u53bb\u566a\u3002\u901a\u8fc7\u8bad\u7ec3\u4e00\u4e2a\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\uff0cCNN\u53ef\u4ee5\u5b66\u4e60\u56fe\u50cf\u53bb\u566a\u7684\u590d\u6742\u6a21\u5f0f\u548c\u7279\u5f81\u3002<\/p>\n<\/p>\n
\u5b9e\u73b0\u65b9\u6cd5<\/h4>\n<\/p>\n
\u4f7f\u7528TensorFlow\u6216PyTorch\u7b49\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u53ef\u4ee5\u5b9e\u73b0\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u53bb\u566a\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u4f8b\u5b50\uff0c\u4f7f\u7528TensorFlow\u5b9e\u73b0\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u53bb\u566a\uff1a<\/p>\n<\/p>\n
import tensorflow as tf<\/p>\n
from tensorflow.keras.models import Model<\/p>\n
from tensorflow.keras.layers import Input, Conv2D, MaxPooling2D, UpSampling2D<\/p>\n
import numpy as np<\/p>\n
import cv2<\/p>\n
\u6784\u5efa\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u6a21\u578b<\/strong><\/h2>\n
input_img = Input(shape=(None, None, 1))<\/p>\n
x = Conv2D(64, (3, 3), activation='relu', padding='same')(input_img)<\/p>\n
x = MaxPooling2D((2, 2), padding='same')(x)<\/p>\n
x = Conv2D(64, (3, 3), activation='relu', padding='same')(x)<\/p>\n
encoded = MaxPooling2D((2, 2), padding='same')(x)<\/p>\n
x = Conv2D(64, (3, 3), activation='relu', padding='same')(encoded)<\/p>\n
x = UpSampling2D((2, 2))(x)<\/p>\n
x = Conv2D(64, (3, 3), activation='relu', padding='same')(x)<\/p>\n
x = UpSampling2D((2, 2))(x)<\/p>\n
decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same')(x)<\/p>\n
autoencoder = Model(input_img, decoded)<\/p>\n
autoencoder.compile(optimizer='adam', loss='binary_crossentropy')<\/p>\n
\u52a0\u8f7d\u8bad\u7ec3\u6570\u636e\uff08\u8fd9\u91cc\u4f7f\u7528MNIST\u6570\u636e\u96c6\u4f5c\u4e3a\u793a\u4f8b\uff09<\/strong><\/h2>\n
(x_train, _), (x_test, _) = tf.keras.datasets.mnist.load_data()<\/p>\n
x_train = x_train.astype('float32') \/ 255.<\/p>\n
x_test = x_test.astype('float32') \/ 255.<\/p>\n
x_train = np.reshape(x_train, (len(x_train), 28, 28, 1))<\/p>\n
x_test = np.reshape(x_test, (len(x_test), 28, 28, 1))<\/p>\n
\u6dfb\u52a0\u566a\u58f0<\/strong><\/h2>\n
noise_factor = 0.5<\/p>\n
x_train_noisy = x_train + noise_factor * np.random.normal(loc=0.0, scale=1.0, size=x_train.shape)<\/p>\n
x_test_noisy = x_test + noise_factor * np.random.normal(loc=0.0, scale=1.0, size=x_test.shape)<\/p>\n
x_train_noisy = np.clip(x_train_noisy, 0., 1.)<\/p>\n
x_test_noisy = np.clip(x_test_noisy, 0., 1.)<\/p>\n
\u8bad\u7ec3\u6a21\u578b<\/strong><\/h2>\n
autoencoder.fit(x_train_noisy, x_train,<\/p>\n
                epochs=100,<\/p>\n
                batch_size=128,<\/p>\n
                shuffle=True,<\/p>\n
                validation_data=(x_test_noisy, x_test))<\/p>\n
\u53bb\u566a<\/strong><\/h2>\n
denoised_images = autoencoder.predict(x_test_noisy)<\/p>\n
\u663e\u793a\u7ed3\u679c<\/strong><\/h2>\n
import matplotlib.pyplot as plt<\/p>\n
n = 10<\/p>\n
plt.figure(figsize=(20, 4))<\/p>\n
for i in range(n):<\/p>\n
    ax = plt.subplot(3, n, i + 1)<\/p>\n
    plt.imshow(x_test_noisy[i].reshape(28, 28), cmap='gray')<\/p>\n
    plt.title("Noisy")<\/p>\n
    plt.axis('off')<\/p>\n
    ax = plt.subplot(3, n, i + 1 + n)<\/p>\n
    plt.imshow(x_test[i].reshape(28, 28), cmap='gray')<\/p>\n
    plt.title("Original")<\/p>\n
    plt.axis('off')<\/p>\n
    ax = plt.subplot(3, n, i + 1 + 2 * n)<\/p>\n
    plt.imshow(denoised_images[i].reshape(28, 28), cmap='gray')<\/p>\n
    plt.title("Denoised")<\/p>\n
    plt.axis('off')<\/p>\n
plt.show()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u5728\u8fd9\u4e2a\u4f8b\u5b50\u4e2d\uff0c\u6211\u4eec\u6784\u5efa\u4e86\u4e00\u4e2a\u7b80\u5355\u7684\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\uff0c\u5e76\u5728MNIST\u6570\u636e\u96c6\u4e0a\u8fdb\u884c\u8bad\u7ec3\u3002\u6211\u4eec\u6dfb\u52a0\u566a\u58f0\u5230\u8bad\u7ec3\u6570\u636e\u4e2d\uff0c\u7136\u540e\u4f7f\u7528\u8bad\u7ec3\u597d\u7684\u6a21\u578b\u53bb\u566a\u3002<\/p>\n<\/p>\n
\u603b\u7ed3<\/h3>\n<\/p>\n
\u4f7f\u7528Python\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a\u7684\u65b9\u6cd5\u6709\u5f88\u591a\uff0c\u5305\u62ec\u5747\u503c\u6ee4\u6ce2\u3001\u4e2d\u503c\u6ee4\u6ce2\u3001\u9ad8\u65af\u6ee4\u6ce2\u3001\u53cc\u8fb9\u6ee4\u6ce2\u3001\u9891\u57df\u6ee4\u6ce2\u3001\u975e\u5c40\u90e8\u5747\u503c\u6ee4\u6ce2\u548c\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u53bb\u566a<\/strong>\u7b49\u3002\u6bcf\u79cd\u65b9\u6cd5\u90fd\u6709\u5176\u4f18\u70b9\u548c\u7f3a\u70b9\uff0c\u9009\u62e9\u5408\u9002\u7684\u65b9\u6cd5\u53ef\u4ee5\u6839\u636e\u5177\u4f53\u5e94\u7528\u573a\u666f\u548c\u56fe\u50cf\u7279\u70b9\u6765\u51b3\u5b9a\u3002\u5728\u5b9e\u9645\u5e94\u7528\u4e2d\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u591a\u79cd\u65b9\u6cd5\uff0c\u5e76\u6839\u636e\u53bb\u566a\u6548\u679c\u9009\u62e9\u6700\u4f73\u65b9\u6848\u3002<\/p>\n<\/p>\n
\u76f8\u5173\u95ee\u7b54FAQs\uff1a<\/strong><\/h2>\n
 \u5982\u4f55\u9009\u62e9\u5408\u9002\u7684\u53bb\u566a\u7b97\u6cd5\uff1f<\/strong>
\u9009\u62e9\u53bb\u566a\u7b97\u6cd5\u65f6\uff0c\u9700\u8003\u8651\u56fe\u50cf\u7684\u7279\u5f81\u548c\u566a\u58f0\u7c7b\u578b\u3002\u5e38\u89c1\u7684\u53bb\u566a\u7b97\u6cd5\u6709\u5747\u503c\u6ee4\u6ce2\u3001\u4e2d\u503c\u6ee4\u6ce2\u3001\u53cc\u8fb9\u6ee4\u6ce2\u548c\u5c0f\u6ce2\u53d8\u6362\u7b49\u3002\u5747\u503c\u6ee4\u6ce2\u9002\u5408\u5904\u7406\u9ad8\u65af\u566a\u58f0\uff0c\u800c\u4e2d\u503c\u6ee4\u6ce2\u5728\u53bb\u9664\u6912\u76d0\u566a\u58f0\u65f6\u8868\u73b0\u4f18\u5f02\u3002\u53cc\u8fb9\u6ee4\u6ce2\u517c\u987e\u53bb\u566a\u548c\u4fdd\u7559\u8fb9\u7f18\uff0c\u9002\u5408\u9700\u8981\u4fdd\u6301\u7ec6\u8282\u7684\u56fe\u50cf\u3002\u7528\u6237\u53ef\u6839\u636e\u5177\u4f53\u9700\u6c42\u548c\u566a\u58f0\u60c5\u51b5\u8fdb\u884c\u9009\u62e9\u3002<\/p>\n
\u5982\u4f55\u5728Python\u4e2d\u5b9e\u73b0\u4e2d\u503c\u6ee4\u6ce2\uff1f<\/strong>
\u5728Python\u4e2d\u5b9e\u73b0\u4e2d\u503c\u6ee4\u6ce2\u975e\u5e38\u7b80\u5355\uff0c\u901a\u5e38\u4f7f\u7528OpenCV\u5e93\u3002\u53ef\u4ee5\u901a\u8fc7\u4ee5\u4e0b\u4ee3\u7801\u5b9e\u73b0\u4e2d\u503c\u6ee4\u6ce2\uff1a  <\/p>\n
import cv2\n\n# \u8bfb\u53d6\u7070\u5ea6\u56fe\u50cf\nimage = cv2.imread('image_path', cv2.IMREAD_GRAYSCALE)\n\n# \u5e94\u7528\u4e2d\u503c\u6ee4\u6ce2\ndenoised_image = cv2.medianBlur(image, 5)\n\n# \u4fdd\u5b58\u6216\u5c55\u793a\u53bb\u566a\u540e\u7684\u56fe\u50cf\ncv2.imwrite('denoised_image_path', denoised_image)\n<\/code><\/pre>\n
\u4e0a\u8ff0\u4ee3\u7801\u5c06\u5bf9\u56fe\u50cf\u5e94\u7528\u5927\u5c0f\u4e3a5\u7684\u4e2d\u503c\u6ee4\u6ce2\u5668\uff0c\u6709\u6548\u53bb\u9664\u566a\u58f0\u3002<\/p>\n
\u4f7f\u7528\u53bb\u566a\u6280\u672f\u65f6\u4f1a\u5f71\u54cd\u56fe\u50cf\u8d28\u91cf\u5417\uff1f<\/strong>
\u53bb\u566a\u6280\u672f\u5728\u6709\u6548\u53bb\u9664\u566a\u58f0\u7684\u540c\u65f6\uff0c\u53ef\u80fd\u4f1a\u5bf9\u56fe\u50cf\u7684\u7ec6\u8282\u548c\u7eb9\u7406\u9020\u6210\u4e00\u5b9a\u5f71\u54cd\u3002\u4f8b\u5982\uff0c\u8fc7\u5ea6\u53bb\u566a\u53ef\u80fd\u5bfc\u81f4\u56fe\u50cf\u6a21\u7cca\u6216\u7ec6\u8282\u4e22\u5931\u3002\u56e0\u6b64\uff0c\u9009\u62e9\u5408\u9002\u7684\u53bb\u566a\u7b97\u6cd5\u548c\u53c2\u6570\u8bbe\u7f6e\u81f3\u5173\u91cd\u8981\u3002\u7528\u6237\u53ef\u4ee5\u901a\u8fc7\u8c03\u6574\u6ee4\u6ce2\u5668\u7684\u5927\u5c0f\u6216\u9009\u62e9\u4e0d\u540c\u7684\u53bb\u566a\u65b9\u6cd5\u6765\u5e73\u8861\u566a\u58f0\u53bb\u9664\u548c\u56fe\u50cf\u8d28\u91cf\u4e4b\u95f4\u7684\u5173\u7cfb\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"\u5982\u4f55\u7528Python\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a \u4f7f\u7528Python\u7ed9\u7070\u5ea6\u56fe\u50cf\u53bb\u566a\u7684\u65b9\u6cd5\u5305\u62ec\u5747\u503c\u6ee4\u6ce2\u3001\u4e2d\u503c\u6ee4\u6ce2\u3001\u9ad8\u65af\u6ee4\u6ce2\u3001\u53cc\u8fb9\u6ee4 […]","protected":false},"author":3,"featured_media":1115821,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[37],"tags":[],"acf":[],"_links":{"self":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1115807"}],"collection":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/comments?post=1115807"}],"version-history":[{"count":"1","href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1115807\/revisions"}],"predecessor-version":[{"id":1115823,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1115807\/revisions\/1115823"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media\/1115821"}],"wp:attachment":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media?parent=1115807"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/categories?post=1115807"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/tags?post=1115807"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}