图像锐化，阈值

设置阈值，大于T为梯度，否则不变。即：

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

def gradient_T(img, T):
	result = np.array(img)
	for i in range(img.shape[0]-1):
		for j in range(img.shape[1]-1):
			g = np.abs(int(img[i][j]) - int(img[i+1][j])) + np.abs(int(img[i][j]) - int(img[i][j+1]))
			if img[i][j] > T:
				result[i][j] = g
	return result


img = cv.imread('lena512color.tiff', 0)
result1 = gradient_T(img, 20)
result2 = gradient_T(img, 80)
result3 = gradient_T(img, 150)


res1 = np.hstack((img, result1))
res2 = np.hstack((result2, result3))
cv.imshow('img and result1', res1)
cv.imshow('result2 and result3', res2)

cv.waitKey(0)

第一张是原图，第二张是T=20时的梯度图像，第三张是T=80时的梯度图像，最后一张是T=150时的梯度图像，由图可以看出，当阈值越大，轮廓越不明显。

为边缘处为一个特定的灰度级。即

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

def gradient_T(img, T):
	result = np.array(img)
	for i in range(img.shape[0]-1):
		for j in range(img.shape[1]-1):
			g = np.abs(int(img[i][j]) - int(img[i+1][j])) + np.abs(int(img[i][j]) - int(img[i][j+1]))
			if g > T:
				result[i][j] = 80
	return result


img = cv.imread('lena512color.tiff', 0)
result1 = gradient_T(img, 20)


res1 = np.hstack((img, result1))
cv.imshow('img and result1', res1)

cv.waitKey(0)