Canny边缘检测:第二步——边缘细化
在这里我们完成Canny边缘检测的第三步。
我们从在前面的文章中求出的边缘梯度进行非极大值抑制,来对边缘进行细化。
非极大值抑制是对除去非极大值以外的值的操作的总称(这个术语在其它的任务中也经常出现)。
在这里,我们比较我们我们所关注的地方梯度的法线方向邻接的三个像素点的梯度幅值,如果该点的梯度值不比其它两个像素大,那么这个地方的值设置为0。
也就是说,我们在注意梯度幅值\text{edge}(x,y)的时候,可以根据下式由梯度方向\text{angle}(x,y)来变换\text{edge}(x,y):
- $$\text{angle}(x,y) = 0$$
如果在$$\text{edge}(x,y)$$、$$\text{edge}(x-1,y)$$、$$\text{edge}(x+1,y)$$中$$\text{edge}(x,y)$$不是最大的,那么$$\text{edge}(x,y)=0$$;
-
$$\text{angle}(x,y) = 45$$
如果在$$\text{edge}(x,y)$$、$$\text{edge}(x-1,y)$$、$$\text{edge}(x+1,y)$$中$$\text{edge}(x,y)$$不是最大的,那么$$\text{edge}(x,y)=0$$;
-
$$\text{angle}(x,y) = 90$$
如果在$$\text{edge}(x,y)$$、$$\text{edge}(x-1,y)$$、$$\text{edge}(x+1,y)$$中$$\text{edge}(x,y)$$不是最大的,那么$$\text{edge}(x,y)=0$$;
-
$$\text{angle}(x,y) = 135$$
如果在$$\text{edge}(x,y)$$、$$\text{edge}(x-1,y)$$、$$\text{edge}(x+1,y)$$中$$\text{edge}(x,y)$$不是最大的,那么$$\text{edge}(x,y)=0$$;
python实现:
import cv2
import numpy as np
import matplotlib.pyplot as plt
def Canny_step2(img):
# Gray scale
def BGR2GRAY(img):
b = img[:, :, 0].copy()
g = img[:, :, 1].copy()
r = img[:, :, 2].copy()
# Gray scale
out = 0.2126 * r + 0.7152 * g + 0.0722 * b
out = out.astype(np.uint8)
return out
# Gaussian filter for grayscale
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 /= (sigma * np.sqrt(2 * np.pi))
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]
out = out.astype(np.uint8)
out = out[..., 0]
return out
# sobel filter
def sobel_filter(img, K_size=3):
if len(img.shape) == 3:
H, W, C = img.shape
else:
H, W = img.shape
# Zero padding
pad = K_size // 2
out = np.zeros((H + pad * 2, W + pad * 2), dtype=np.float)
out[pad: pad + H, pad: pad + W] = img.copy().astype(np.float)
tmp = out.copy()
out_v = out.copy()
out_h = out.copy()
## Sobel vertical
Kv = [[1., 2., 1.],[0., 0., 0.], [-1., -2., -1.]]
## Sobel horizontal
Kh = [[1., 0., -1.],[2., 0., -2.],[1., 0., -1.]]
# filtering
for y in range(H):
for x in range(W):
out_v[pad + y, pad + x] = np.sum(Kv * (tmp[y: y + K_size, x: x + K_size]))
out_h[pad + y, pad + x] = np.sum(Kh * (tmp[y: y + K_size, x: x + K_size]))
out_v = np.clip(out_v, 0, 255)
out_h = np.clip(out_h, 0, 255)
out_v = out_v[pad: pad + H, pad: pad + W].astype(np.uint8)
out_h = out_h[pad: pad + H, pad: pad + W].astype(np.uint8)
return out_v, out_h
def get_edge_angle(fx, fy):
# get edge strength
edge = np.sqrt(np.power(fx.astype(np.float32), 2) + np.power(fy.astype(np.float32), 2))
edge = np.clip(edge, 0, 255)
fx = np.maximum(fx, 1e-5)
#fx[np.abs(fx) <= 1e-5] = 1e-5
# get edge angle
angle = np.arctan(fy / fx)
return edge, angle
def angle_quantization(angle):
angle = angle / np.pi * 180
angle[angle < -22.5] = 180 + angle[angle < -22.5]
_angle = np.zeros_like(angle, dtype=np.uint8)
_angle[np.where(angle <= 22.5)] = 0
_angle[np.where((angle > 22.5) & (angle <= 67.5))] = 45
_angle[np.where((angle > 67.5) & (angle <= 112.5))] = 90
_angle[np.where((angle > 112.5) & (angle <= 157.5))] = 135
return _angle
def non_maximum_suppression(angle, edge):
H, W = angle.shape
_edge = edge.copy()
for y in range(H):
for x in range(W):
if angle[y, x] == 0:
dx1, dy1, dx2, dy2 = -1, 0, 1, 0
elif angle[y, x] == 45:
dx1, dy1, dx2, dy2 = -1, 1, 1, -1
elif angle[y, x] == 90:
dx1, dy1, dx2, dy2 = 0, -1, 0, 1
elif angle[y, x] == 135:
dx1, dy1, dx2, dy2 = -1, -1, 1, 1
if x == 0:
dx1 = max(dx1, 0)
dx2 = max(dx2, 0)
if x == W-1:
dx1 = min(dx1, 0)
dx2 = min(dx2, 0)
if y == 0:
dy1 = max(dy1, 0)
dy2 = max(dy2, 0)
if y == H-1:
dy1 = min(dy1, 0)
dy2 = min(dy2, 0)
if max(max(edge[y, x], edge[y + dy1, x + dx1]), edge[y + dy2, x + dx2]) != edge[y, x]:
_edge[y, x] = 0
return _edge
# grayscale
gray = BGR2GRAY(img)
# gaussian filtering
gaussian = gaussian_filter(gray, K_size=5, sigma=1.4)
# sobel filtering
fy, fx = sobel_filter(gaussian, K_size=3)
# get edge strength, angle
edge, angle = get_edge_angle(fx, fy)
# angle quantization
angle = angle_quantization(angle)
# non maximum suppression
edge = non_maximum_suppression(angle, edge)
return edge, angle
# Read image
img = cv2.imread("imori.jpg").astype(np.float32)
# Canny (step2)
edge, angle = Canny_step2(img)
edge = edge.astype(np.uint8)
angle = angle.astype(np.uint8)
# Save result
cv2.imwrite("out.jpg", edge)
cv2.imshow("result", edge)
cv2.imwrite("out2.jpg", angle)
cv2.imshow("result2", angle)
cv2.waitKey(0)
cv2.destroyAllWindows()
c++实现:
#include <opencv2/core.hpp>
#include <opencv2/highgui.hpp>
#include <iostream>
#include <math.h>
// RGB to Gray scale
cv::Mat BGR2GRAY(cv::Mat img){
// get height and width
int height = img.rows;
int width = img.cols;
int channel = img.channels();
// prepare output
cv::Mat out = cv::Mat::zeros(height, width, CV_8UC1);
// BGR -> Gray
for (int y = 0; y < height; y++){
for (int x = 0; x < width; x++){
out.at<uchar>(y, x) = (int)((float)img.at<cv::Vec3b>(y, x)[0] * 0.0722 + \
(float)img.at<cv::Vec3b>(y, x)[1] * 0.7152 + \
(float)img.at<cv::Vec3b>(y, x)[2] * 0.2126);
}
}
return out;
}
float clip(float value, float min, float max){
return fmin(fmax(value, 0), 255);
}
// gaussian filter
cv::Mat gaussian_filter(cv::Mat img, double sigma, int kernel_size){
int height = img.rows;
int width = img.cols;
int channel = img.channels();
// prepare output
cv::Mat out = cv::Mat::zeros(height, width, CV_8UC3);
if (channel == 1) {
out = cv::Mat::zeros(height, width, CV_8UC1);
}
// prepare kernel
int pad = floor(kernel_size / 2);
int _x = 0, _y = 0;
double kernel_sum = 0;
// get gaussian kernel
float kernel[kernel_size][kernel_size];
for (int y = 0; y < kernel_size; y++){
for (int x = 0; x < kernel_size; x++){
_y = y - pad;
_x = x - pad;
kernel[y][x] = 1 / (2 * M_PI * sigma * sigma) * exp( - (_x * _x + _y * _y) / (2 * sigma * sigma));
kernel_sum += kernel[y][x];
}
}
for (int y = 0; y < kernel_size; y++){
for (int x = 0; x < kernel_size; x++){
kernel[y][x] /= kernel_sum;
}
}
// filtering
double v = 0;
for (int y = 0; y < height; y++){
for (int x = 0; x < width; x++){
// for BGR
if (channel == 3){
for (int c = 0; c < channel; c++){
v = 0;
for (int dy = -pad; dy < pad + 1; dy++){
for (int dx = -pad; dx < pad + 1; dx++){
if (((x + dx) >= 0) && ((y + dy) >= 0) && ((x + dx) < width) && ((y + dy) < height)){
v += (double)img.at<cv::Vec3b>(y + dy, x + dx)[c] * kernel[dy + pad][dx + pad];
}
}
}
out.at<cv::Vec3b>(y, x)[c] = (uchar)clip(v, 0, 255);
}
} else {
// for Gray
v = 0;
for (int dy = -pad; dy < pad + 1; dy++){
for (int dx = -pad; dx < pad + 1; dx++){
if (((x + dx) >= 0) && ((y + dy) >= 0) && ((x + dx) < width) && ((y + dy) < height)){
v += (double)img.at<uchar>(y + dy, x + dx) * kernel[dy + pad][dx + pad];
}
}
}
out.at<uchar>(y, x) = (uchar)clip(v, 0, 255);
}
}
}
return out;
}
// Sobel filter
cv::Mat sobel_filter(cv::Mat img, int kernel_size, bool horizontal){
int height = img.rows;
int width = img.cols;
int channel = img.channels();
// prepare output
cv::Mat out = cv::Mat::zeros(height, width, CV_8UC1);
// prepare kernel
double kernel[kernel_size][kernel_size] = {{1, 2, 1}, {0, 0, 0}, {-1, -2, -1}};
if (horizontal){
kernel[0][1] = 0;
kernel[0][2] = -1;
kernel[1][0] = 2;
kernel[1][2] = -2;
kernel[2][0] = 1;
kernel[2][1] = 0;
}
int pad = floor(kernel_size / 2);
double v = 0;
// filtering
for (int y = 0; y < height; y++){
for (int x = 0; x < width; x++){
v = 0;
for (int dy = -pad; dy < pad + 1; dy++){
for (int dx = -pad; dx < pad + 1; dx++){
if (((y + dy) >= 0) && (( x + dx) >= 0) && ((y + dy) < height) && ((x + dx) < width)){
v += (double)img.at<uchar>(y + dy, x + dx) * kernel[dy + pad][dx + pad];
}
}
}
out.at<uchar>(y, x) = (uchar)clip(v, 0, 255);
}
}
return out;
}
// get edge
cv::Mat get_edge(cv::Mat fx, cv::Mat fy){
// get height and width
int height = fx.rows;
int width = fx.cols;
// prepare output
cv::Mat out = cv::Mat::zeros(height, width, CV_8UC1);
double _fx, _fy;
for(int y = 0; y < height; y++){
for(int x = 0; x < width; x++){
_fx = (double)fx.at<uchar>(y, x);
_fy = (double)fy.at<uchar>(y, x);
out.at<uchar>(y, x) = (uchar)clip(sqrt(_fx * _fx + _fy * _fy), 0, 255);
}
}
return out;
}
// get angle
cv::Mat get_angle(cv::Mat fx, cv::Mat fy){
// get height and width
int height = fx.rows;
int width = fx.cols;
// prepare output
cv::Mat out = cv::Mat::zeros(height, width, CV_8UC1);
double _fx, _fy;
double angle;
for(int y = 0; y < height; y++){
for(int x = 0; x < width; x++){
_fx = fmax((double)fx.at<uchar>(y, x), 0.000001);
_fy = (double)fy.at<uchar>(y, x);
angle = atan2(_fy, _fx);
angle = angle / M_PI * 180;
if(angle < -22.5){
angle = 180 + angle;
} else if (angle >= 157.5) {
angle = angle - 180;
}
//std::cout << angle << " " ;
// quantization
if (angle <= 22.5){
out.at<uchar>(y, x) = 0;
} else if (angle <= 67.5){
out.at<uchar>(y, x) = 45;
} else if (angle <= 112.5){
out.at<uchar>(y, x) = 90;
} else {
out.at<uchar>(y, x) = 135;
}
}
}
return out;
}
// non maximum suppression
cv::Mat non_maximum_suppression(cv::Mat angle, cv::Mat edge){
int height = angle.rows;
int width = angle.cols;
int channel = angle.channels();
int dx1, dx2, dy1, dy2;
int now_angle;
// prepare output
cv::Mat _edge = cv::Mat::zeros(height, width, CV_8UC1);
for (int y = 0; y < height; y++){
for (int x = 0; x < width; x++){
now_angle = angle.at<uchar>(y, x);
// angle condition
if (now_angle == 0){
dx1 = -1;
dy1 = 0;
dx2 = 1;
dy2 = 0;
} else if(now_angle == 45) {
dx1 = -1;
dy1 = 1;
dx2 = 1;
dy2 = -1;
} else if(now_angle == 90){
dx1 = 0;
dy1 = -1;
dx2 = 0;
dy2 = 1;
} else {
dx1 = -1;
dy1 = -1;
dx2 = 1;
dy2 = 1;
}
if (x == 0){
dx1 = fmax(dx1, 0);
dx2 = fmax(dx2, 0);
}
if (x == (width - 1)){
dx1 = fmin(dx1, 0);
dx2 = fmin(dx2, 0);
}
if (y == 0){
dy1 = fmax(dy1, 0);
dy2 = fmax(dy2, 0);
}
if (y == (height - 1)){
dy1 = fmin(dy1, 0);
dy2 = fmin(dy2, 0);
}
// if pixel is max among adjuscent pixels, pixel is kept
if (fmax(fmax(edge.at<uchar>(y, x), edge.at<uchar>(y + dy1, x + dx1)), edge.at<uchar>(y + dy2, x + dx2)) == edge.at<uchar>(y, x)) {
_edge.at<uchar>(y, x) = edge.at<uchar>(y, x);
}
}
}
return _edge;
}
// Canny step 2
int Canny_step2(cv::Mat img){
// BGR -> Gray
cv::Mat gray = BGR2GRAY(img);
// gaussian filter
cv::Mat gaussian = gaussian_filter(gray, 1.4, 5);
// sobel filter (vertical)
cv::Mat fy = sobel_filter(gaussian, 3, false);
// sobel filter (horizontal)
cv::Mat fx = sobel_filter(gaussian, 3, true);
// get edge
cv::Mat edge = get_edge(fx, fy);
// get angle
cv::Mat angle = get_angle(fx, fy);
// edge non-maximum suppression
edge = non_maximum_suppression(angle, edge);
//cv::imwrite("out.jpg", out);
cv::imshow("answer(edge)", edge);
cv::imshow("answer(angle)", angle);
cv::waitKey(0);
cv::destroyAllWindows();
return 0;
}
int main(int argc, const char* argv[]){
// read image
cv::Mat img = cv::imread("imori.jpg", cv::IMREAD_COLOR);
// Canny step 2
Canny_step2(img);
return 0;
}
输入:
输出: