Skip to content

Instantly share code, notes, and snippets.

@YutaSeya

YutaSeya/main.cpp Secret

Last active January 9, 2020 08:07
Show Gist options
  • Save YutaSeya/05b52dc8fe3328c13e98bb7ea89f7039 to your computer and use it in GitHub Desktop.
Save YutaSeya/05b52dc8fe3328c13e98bb7ea89f7039 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
main.cpp
#include <opencv2/opencv.hpp>
#include <iostream>
#include <cmath>
#include "myfilter.h"
using namespace cv;
using namespace std;
//#define __CHECK_OPENCV
//#define __CHECK_AVERAGE
//#define __CHECK_GAUSSIAN
#define __CHECK_SOBEL
#define __CHECK_LAPLACIAN
int main()
{
Mat data;
data = imread("3.jpg");
imshow("read data", data);
#if __CHECK_AVERAGE 1
Mat ave;
ave = myAverageFilter(data, Size(3, 3));
imshow("my average filter", ave);
#endif
#if __CHECK_GAUSSIAN 1
Mat gauss;
gauss = myGaussianFilter(data, Size(3, 3), 1.2);
imshow("my gaussian filter", gauss);
#endif
#if __CHECK_SOBEL 1
Mat sobel_h, sobel_v;
sobel_h = mySobelFilter(data, 0, 1);
sobel_v = mySobelFilter(data, 1, 1);
imshow("my sobel horizontal", sobel_h);
imshow("my sobel vertical", sobel_v);
#endif
#if __CHECK_LAPLACIAN 1
Mat laplacian;
Mat laplacian_binary;
laplacian = myLaplacianFilter(data, 1);
imshow("my laplacian", laplacian);
#endif
#if __CHECK_OPENCV 1
Mat cvlap;
Laplacian(data, cvlap, CV_32F, 3);
convertScaleAbs(cvlap, cvlap, 1, 0);
imshow("opencv laplacian", cvlap);
Mat canny;
Canny(data, canny, 50, 200);
imshow("opencv canny", canny);
#endif
waitKey();
return 0;
}
Raw
myfilter.cpp
#include "myfilter.h"
#define PI 3.1415926535f
using namespace cv;
using namespace std;
//#define _DEBUG
/**
* @brief 自作平均フィルタ
* @detail 端の計算処理はしていないため、端の画素は入力画像と同じように表示される。
* 理想は、端のほうも処理をいれることだが、バグフィックスが面倒なためいれていない。
**/
Mat myAverageFilter(Mat data, Size ksize)
{
int height = data.size().height;
int width = data.size().width;
Mat out;
out = data.clone();
uint64_t sumR = 0;
uint64_t sumG = 0;
uint64_t sumB = 0;
for (int x = ksize.width / 2; x < width - ksize.width / 2; x++) {
for (int y = ksize.height / 2; y < height - ksize.height / 2; y++) {
for (int k = -ksize.width / 2; k <= ksize.width / 2; k++) {
for (int l = -ksize.height / 2; l <= ksize.height / 2; l++) {
sumB += data.at<Vec3b>(y + l, x + k)[0];
sumG += data.at<Vec3b>(y + l, x + k)[1];
sumR += data.at<Vec3b>(y + l, x + k)[2];
}
}
out.at<Vec3b>(y - ksize.height / 2, x - ksize.width / 2)[0] = sumB / ksize.area();
out.at<Vec3b>(y - ksize.height / 2, x - ksize.width / 2)[1] = sumG / ksize.area();
out.at<Vec3b>(y - ksize.height / 2, x - ksize.width / 2)[2] = sumR / ksize.area();
sumB = 0;
sumG = 0;
sumR = 0;
}
}
return out;
}
/**
* @brief 自作ガウシアンフィルタ
* @detail 端の計算処理はしていないため、端の画素は入力画像と同じように表示される。
* 理想は、端のほうも処理をいれることだが、バグフィックスが面倒なためいれていない。
**/
Mat myGaussianFilter(Mat data, Size ksize, float sigma)
{
Mat karnel;
karnel = Mat::zeros(ksize, CV_32F);
float gause_sum = 0.0f;
// ガウシアンフィルタ用のカーネルを作成
for (int x = -ksize.width / 2; x <= ksize.width / 2; x++) {
for (int y = -ksize.height / 2; y <= ksize.height / 2; y++) {
karnel.at<float>(y + ksize.height / 2, x + ksize.width / 2) = (float)exp(-((x * x + y * y) / (2 * sigma * sigma))) / (2 * PI * sigma * sigma);
gause_sum += karnel.at<float>(y + ksize.height / 2, x + ksize.width / 2);
}
}
// スカラ倍をして、和が1になるように落とし込む
karnel = karnel / gause_sum;
#if _DEBUG 1
cout << karnel << endl;
#endif
int height = data.size().height;
int width = data.size().width;
Mat out;
out = data.clone();
float sumR = 0;
float sumG = 0;
float sumB = 0;
for (int x = ksize.width / 2; x < width - ksize.width / 2; x++) {
for (int y = ksize.height / 2; y < height - ksize.height / 2; y++) {
for (int k = -ksize.width / 2; k <= ksize.width / 2; k++) {
for (int l = -ksize.height / 2; l <= ksize.height / 2; l++) {
sumB += (float)data.at<Vec3b>(y + l, x + k)[0] * karnel.at<float>(l + ksize.height / 2, k + ksize.width / 2);
sumG += (float)data.at<Vec3b>(y + l, x + k)[1] * karnel.at<float>(l + ksize.height / 2, k + ksize.width / 2);
sumR += (float)data.at<Vec3b>(y + l, x + k)[2] * karnel.at<float>(l + ksize.height / 2, k + ksize.width / 2);
}
}
out.at<Vec3b>(y - ksize.height / 2, x - ksize.width / 2)[0] = sumB;
out.at<Vec3b>(y - ksize.height / 2, x - ksize.width / 2)[1] = sumG;
out.at<Vec3b>(y - ksize.height / 2, x - ksize.width / 2)[2] = sumR;
sumB = 0.0f;
sumG = 0.0f;
sumR = 0.0f;
}
}
return out;
}
Mat mySobelFilter(Mat data, int direction, int color)
{
Mat karnel;
karnel = Mat::zeros(Size(3, 3), CV_32F);
if (direction == 0) {
karnel.at<float>(0, 0) = 1;
karnel.at<float>(0, 2) = -1;
karnel.at<float>(1, 0) = 2;
karnel.at<float>(1, 2) = -2;
karnel.at<float>(2, 0) = 1;
karnel.at<float>(2, 2) = -1;
}
else {
karnel.at<float>(0, 0) = 1;
karnel.at<float>(0, 1) = 2;
karnel.at<float>(0, 2) = 1;
karnel.at<float>(2, 0) = -1;
karnel.at<float>(2, 1) = -2;
karnel.at<float>(2, 2) = -1;
}
#if _DEBUG 1
cout << karnel << endl;
#endif
int height = data.size().height;
int width = data.size().width;
Mat out;
if (color == 0) {
out = data.clone();
float sumR = 0;
float sumG = 0;
float sumB = 0;
for (int x = 1; x < width - 1; x++) {
for (int y = 1; y < height - 1; y++) {
for (int k = -1; k <= 1; k++) {
for (int l = -1; l <= 1; l++) {
sumB += (float)data.at<Vec3b>(y + l, x + k)[0] * karnel.at<float>(l + 1, k + 1);
sumG += (float)data.at<Vec3b>(y + l, x + k)[1] * karnel.at<float>(l + 1, k + 1);
sumR += (float)data.at<Vec3b>(y + l, x + k)[2] * karnel.at<float>(l + 1, k + 1);
}
}
if (sumB < 0.0f) sumB = 0;
if (sumG < 0.0f) sumG = 0;
if (sumR < 0.0f) sumR = 0;
out.at<Vec3b>(y - 1, x - 1)[0] = sumB;
out.at<Vec3b>(y - 1, x - 1)[1] = sumG;
out.at<Vec3b>(y - 1, x - 1)[2] = sumR;
sumB = 0.0f;
sumG = 0.0f;
sumR = 0.0f;
}
}
}
else {
Mat gray_data;
cvtColor(data, gray_data, CV_BGR2GRAY);
out = gray_data.clone();
float sum_data = 0.0f;
for (int x = 1; x < width - 1; x++) {
for (int y = 1; y < height - 1; y++) {
for (int k = -1; k <= 1; k++) {
for (int l = -1; l <= 1; l++) {
sum_data += (float)gray_data.at<uchar>(y + l, x + k) * karnel.at<float>(l + 1, k + 1);
}
}
if (sum_data < 0.0f) sum_data = 0.0f;
out.at<uchar>(y - 1, x - 1) = (uchar)sum_data;
sum_data = 0.0f;
}
}
}
return out;
}
Mat myLaplacianFilter(Mat data, int color)
{
Mat karnel;
karnel = Mat::zeros(Size(3, 3), CV_32F);
for (int i = 0; i < 3; i++) {
for (int j = 0; j < 3; j++) {
karnel.at<float>(i, j) = 1;
}
}
karnel.at<float>(1,1) = -8;
#if _DEBUG 1
cout << karnel << endl;
#endif
int height = data.size().height;
int width = data.size().width;
Mat out;
if (color == 0) {
out = data.clone();
float sumR = 0;
float sumG = 0;
float sumB = 0;
for (int x = 1; x < width - 1; x++) {
for (int y = 1; y < height - 1; y++) {
for (int k = -1; k <= 1; k++) {
for (int l = -1; l <= 1; l++) {
sumB += (float)data.at<Vec3b>(y + l, x + k)[0] * karnel.at<float>(l + 1, k + 1);
sumG += (float)data.at<Vec3b>(y + l, x + k)[1] * karnel.at<float>(l + 1, k + 1);
sumR += (float)data.at<Vec3b>(y + l, x + k)[2] * karnel.at<float>(l + 1, k + 1);
}
}
if (sumB < 0.0f) sumB = 0;
if (sumG < 0.0f) sumG = 0;
if (sumR < 0.0f) sumR = 0;
out.at<Vec3b>(y - 1, x - 1)[0] = sumB;
out.at<Vec3b>(y - 1, x - 1)[1] = sumG;
out.at<Vec3b>(y - 1, x - 1)[2] = sumR;
sumB = 0.0f;
sumG = 0.0f;
sumR = 0.0f;
}
}
}
else {
Mat gray_data;
cvtColor(data, gray_data, CV_BGR2GRAY);
out = gray_data.clone();
float sum_data = 0.0f;
for (int x = 1; x < width - 1; x++) {
for (int y = 1; y < height - 1; y++) {
for (int k = -1; k <= 1; k++) {
for (int l = -1; l <= 1; l++) {
sum_data += (float)gray_data.at<uchar>(y + l, x + k) * karnel.at<float>(l + 1, k + 1);
}
}
if (sum_data < 0.0f) sum_data = 0.0f;
out.at<uchar>(y - 1, x - 1) = (uchar)sum_data;
sum_data = 0.0f;
}
}
}
return out;
}
Raw
myfilter.h
#pragma once
#include <opencv2/opencv.hpp>
#include <iostream>
#include <cmath>
#define _DEBUG
cv::Mat myAverageFilter(cv::Mat data, cv::Size ksize);
cv::Mat myGaussianFilter(cv::Mat data, cv::Size ksize, float sigma);
cv::Mat mySobelFilter(cv::Mat data, int direction, int color);
cv::Mat myLaplacianFilter(cv::Mat data, int color);
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment