梯度方向卷积 自己方法

Posted 2023-03-03 tangjunjun

我也忘记这是什么代码了，几年前的，今天翻到发布出来

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

img = cv.imread(r\'C:\\Users\\51102\\Desktop\\tradition\\1.jpg\',0)
img = cv.cvtColor(img, cv.COLOR_BGR2RGB)
img = np.dot(img[..., :3], [0.299, 0.587, 0.114])
plt.subplot(331)
plt.imshow(img,cmap=\'gray\')
plt.title(\'original_img\')
print(\'original.shape=\',img.shape)

# 构建梯度
gradient_x=np.array([[-1,0,1],[-2,0,2],[-1,0,1]])
gradient_y=np.array([[-1,-2,-1],[0,0,0],[1,2,1]])

plt.subplot(332)

dx=cv.filter2D(img,ddepth=-1,kernel=gradient_x)
print(\'dx.shape=\',dx.shape)
plt.imshow(dx,cmap=\'gray\')
plt.title(\'dx\')
plt.subplot(333)
dy=cv.filter2D(img,-1,gradient_x)
print(\'dy.shape=\',dy.shape)
plt.imshow(dy,cmap=\'gray\')
plt.title(\'dy\')




# 求方向


plt.subplot(334)
magnitude=(np.multiply(dy,dy)+np.multiply(dx,dx))
plt.imshow(magnitude,cmap=\'gray\')
plt.title(\'magnitude\')



cita=cv.phase(dy,dx, angleInDegrees=True)  # 求角度


# 生成高斯滤波器
"""
要生成一个 (2k+1)x(2k+1) 的高斯滤波器，滤波器的各个元素计算公式如下：
H[i, j] = (1/(2*pi*sigma**2))*exp(-1/2*sigma**2((i-k-1)**2 + (j-k-1)**2))
"""
sigma1 = sigma2 = 0.6
gau_sum = 0
gaussian = np.zeros([5, 5])
for i in range(5):
    for j in range(5):
        gaussian[i, j] = math.exp((-1/(2*sigma1*sigma2))*(np.square(i-2-1) + np.square(j-2-1)))/(2*math.pi*sigma1*sigma2)
        gau_sum = gau_sum + gaussian[i, j]
gaussian = gaussian / gau_sum  # 归一化处理

print(\'guss=\',gaussian.shape)

W, H = img.shape  # 原始图片大小 灰度图
_,g=gaussian.shape
new_gray = np.zeros([W-5, H-5])
new_magnitude = np.zeros([W-5, H-5])

for i in range(W-5):
    for j in range(H-5):
        \'\'\'
        滤波处理  我自创一种方法尝试,根据角度，来重新配置权重
        我先降低梯度较大的幅度，对梯度方向进行抑制，其它方向进行加强
        \'\'\'
        guss=gaussian
        g_=int((g-1)/2)

        if 0<cita[i,j]<90 or 180<cita[i,j]<270 or -180<cita[i,j]<-90 or -360<cita[i,j]<-270:
            for k in range(g_):
                for kk in range(g_):
                    guss[g_+1+k,g_+1+kk]=gaussian[g_+1+k,g_+1+kk]/4
                    guss[g_-1 - k, g_ - 1 - kk] = gaussian[g_ - 1 - k, g_ - 1 - kk] / 4
            if 90 < cita[i, j] < 180 or 270 < cita[i, j] < 360 or -90 < cita[i, j] < 0 or -180 < cita[i, j] < -270:
                for k in range(g_):
                    for kk in range(g_):
                        guss[g_ + 1 + k, g_ - 1 - kk] = gaussian[g_ + 1 + k, g_ - 1 - kk] / 4
                        guss[g_ - 1 - k, g_ + 1 + kk] = gaussian[g_ - 1 - k, g_ + 1 + kk] / 4
            new_gray[i, j] = np.sum(img[i:i+5, j:j+5] * guss)
            new_magnitude[i, j] = np.sum(magnitude[i:i + 5, j:j + 5] * guss)








plt.subplot(335)
plt.imshow(new_gray,cmap=\'gray\')
plt.title(\'new_gray\')

plt.subplot(336)
plt.imshow(new_magnitude,cmap=\'gray\')
plt.title(\'new_magnitude\')




plt.show()

梯度方向卷积

 

结果显示：