728x90
728x90
일단 바로 코드부터...
#define _CRT_SECURE_NO_WARNINGS
#include<math.h>
#include<stdio.h>
#include<stdlib.h>
#include <string.h>
#include<windows.h>
double threshold = 0;
void applyPaddingWithAverage(double* paddedImg, int height, int width, int padding, double* origImg);
void calFilter(double* y3, double* py, int fsize, double* filter, int height, int width, int pwidth);
double median(double* arr, int size);
void calMedianFilter(double* y3, double* py, int fsize, int height, int width, int pwidth);
void bilateralFilter(double* paddedImg, double* outputImg, int width, int height, int pwidth, int pheight, int d, double sigmaColor, double sigmaSpace);
void hybridFilter(double* y3, double* py, double* edgey, int fsize, int height, int width, int pwidth, double* gFilter);
void generateGaussianFilter(int n, double sigma, double filter[]);
void upSp(int ratio, double* org, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy);
void nTapInterpolation(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, int n, double* upy);
void makeOutFile(BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, char* fileName, double* y);
void sobelEdgeMap(double* y, int width, int height, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* edgey);
void CumulativeHisto(int height, int width, double* y, int size, int stride, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo);
int main(int argc, char* argv[]) {
BITMAPFILEHEADER bmpFile;
BITMAPINFOHEADER bmpInfo;
FILE* inputFile = NULL;
inputFile = fopen(argv[1], "rb");
fread(&bmpFile, sizeof(BITMAPFILEHEADER), 1, inputFile);
fread(&bmpInfo, sizeof(BITMAPINFOHEADER), 1, inputFile);
// 넓이, 높이 값 받기
int width = bmpInfo.biWidth, height = bmpInfo.biHeight, size = bmpInfo.biSizeImage, bitCnt = bmpInfo.biBitCount, stride = (((bitCnt / 8) * width) + 3) / 4 * 4;
int ratio = 4;
int width2 = width * ratio, height2 = height * ratio, stride2 = (((bitCnt / 8) * width2) + 3) / 4 * 4, size2 = height2 * stride2;
// 이미지 저장 공간
double* y;
y = (double*)calloc(width * height, sizeof(double));
// 이미지 데이터 받아서 저장하기
unsigned char* inputImg = NULL;
inputImg = (unsigned char*)calloc(size, sizeof(unsigned char));
fread(inputImg, sizeof(unsigned char), size, inputFile);
// input
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
y[j * width + i] = inputImg[j * stride + 3 * i];
}
}
//padding
int padding = 2, pwidth = width + padding * 2, pheight = height + padding * 2, psize = pwidth * pheight;
double* py;
py = (double*)calloc(pwidth * pheight, sizeof(double));
applyPaddingWithAverage(py, height, width, padding, y);
// 필터 결과값 저장 공간
double* y3;
y3 = (double*)calloc(width * height, sizeof(double));
// filter
int fsize = padding * 2 + 1;
// Gussian Filter
double G5filter[] = { 1 / 273.0, 4 / 273.0, 7 / 273.0, 4 / 273.0, 1 / 273.0, 4 / 273.0, 16 / 273.0, 26 / 273.0, 16 / 273.0, 4 / 273.0, 7 / 273.0, 26 / 273.0, 41 / 273.0, 26 / 273.0, 7 / 273.0, 4 / 273.0, 16 / 273.0, 26 / 273.0, 16 / 273.0, 4 / 273.0,1 / 273.0, 4 / 273.0, 7 / 273.0, 4 / 273.0, 1 / 273.0 };
// 양방향 필터
bilateralFilter(py, y3, width, height, pwidth, pheight, fsize, 50.0, 10); // 양방향 필터
// 최종 데이터 저장 공간
double* upy;
upy = (double*)calloc(width2 * height2, sizeof(double));
// 업스케일링
upSp(ratio, y3, bmpFile, bmpInfo, upy); // 업스케일링
// 업스케일링에 맞춰 헤드 바꿔주기
bmpInfo.biWidth = height2;
bmpInfo.biHeight = width2;
bmpInfo.biSizeImage = height2 * stride2;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + size2;
double* edgey2;
edgey2 = (double*)calloc(width2 * height2, sizeof(double));
int padding2 = 2, pwidth2 = width2 + padding2 * 2, pheight2 = height2 + padding2 * 2, psize2 = pwidth2 * pheight2;
double* py2;
py2 = (double*)calloc(pwidth2 * pheight2, sizeof(double));
int fsize2 = padding2 * 2 + 1;
// 양방향 필터
applyPaddingWithAverage(py2, height2, width2, padding2, upy);
bilateralFilter(py2, upy, width2, height2, pwidth2, pheight2, fsize2, 60.0, 10.0);
// edge에 따른 필터 적용
applyPaddingWithAverage(py2, height2, width2, padding2, upy);
sobelEdgeMap(upy, width2, height2, bmpFile, bmpInfo, edgey2);
hybridFilter(upy, py2, edgey2, fsize2, height2, width2, pwidth2, G5filter);
int padding3 = 1, pwidth3 = width2 + padding3 * 2, pheight3 = height2 + padding3 * 2, psize3 = pwidth3 * pheight3;
double* py3;
py3 = (double*)calloc(pwidth3 * pheight3, sizeof(double));
int fsize3 = padding3 * 2 + 1;
double* gfilter;
gfilter = (double*)calloc(fsize3 * fsize3, sizeof(double));
// edge에 따른 필터 1회 더 적용
generateGaussianFilter(fsize3, fsize3 / 6.0, gfilter);
applyPaddingWithAverage(py3, height2, width2, padding3, upy);
sobelEdgeMap(upy, width2, height2, bmpFile, bmpInfo, edgey2);
hybridFilter(upy, py3, edgey2, fsize3, height2, width2, pwidth3, gfilter);
// median 필터 적용
applyPaddingWithAverage(py2, height2, width2, padding2, upy);
calMedianFilter(upy, py2, fsize2, height2, width2, pwidth2);
//사진 출력
makeOutFile(bmpFile, bmpInfo, "____.bmp", upy);
//메모리 해제
free(inputImg);
free(y);
free(py);
free(y3);
free(upy);
free(edgey2);
free(py2);
free(py3);
free(gfilter);
fclose(inputFile);
return 0;
}
void calFilter(double* y3, double* py, int fsize, double* filter, int height, int width, int pwidth) {
double sumf = 0;
//각 픽셀에 대해 필터 적용
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
sumf = 0;
// 필터 크기만큼 주변 픽셀 더하기
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
sumf += filter[k * fsize + l] * py[(j + k) * pwidth + (i + l)];
}
}
// 필터 계산 내용 저장
y3[j * width + i] = sumf;
}
}
}
//원래는 히스토그램까지 출력하지만 여기선 필요 없으니 임계값만 정한다.
void CumulativeHisto(int height, int width, double* y, int size, int stride, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo) {
double* histogram;
histogram = (double*)calloc(256, sizeof(double));
// get Histogram
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
histogram[(unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]))]++;
}
}
// 누적 히스토그램 계산
for (int i = 0; i < 255; i++) {
histogram[i + 1] += histogram[i];
}
// 임계값 결정
for (int i = 0; i < 255; i++) {
if (histogram[i] > height * width * 0.985) {
threshold = i;
break;
}
if (i == 254) threshold = 254;
}
free(histogram);
}
// 평균 패딩
void applyPaddingWithAverage(double* paddedImg, int height, int width, int padding, double* origImg) {
int pheight = height + 2 * padding;
int pwidth = width + 2 * padding;
// 원본 이미지 데이터를 중앙에 배치
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
paddedImg[(j + padding) * pwidth + (i + padding)] = origImg[j * width + i];
}
}
// 상하좌우 패딩에 평균 값 적용
for (int j = 0; j < pheight; j++) {
for (int i = 0; i < pwidth; i++) {
// 좌측 패딩 처리
if (i < padding) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = j + k; // 현재 행에서 상하로 2개씩
if (idx >= padding && idx < height + padding) { // 범위 체크
sum += paddedImg[idx * pwidth + padding];
count++;
}
}
paddedImg[j * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
// 우측 패딩 처리
if (i >= width + padding) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = j + k; // 현재 행에서 상하로 2개씩
if (idx >= padding && idx < height + padding) { // 범위 체크
sum += paddedImg[idx * pwidth + (width + padding - 1)];
count++;
}
}
paddedImg[j * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
// 상단 패딩 처리
for (int p = 0; p < padding; p++) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = i + k; // 현재 컬럼에서 좌우로 2개씩
if (idx >= padding && idx < width + padding) { // 범위 체크
sum += paddedImg[padding * pwidth + idx];
count++;
}
}
paddedImg[p * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
// 하단 패딩 처리
for (int p = 0; p < padding; p++) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = i + k; // 현재 컬럼에서 좌우로 2개씩
if (idx >= padding && idx < width + padding) { // 범위 체크
sum += paddedImg[(height + padding - 1) * pwidth + idx];
count++;
}
}
paddedImg[(height + padding + p) * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
}
}
}
// 업셈플링 정하는 곳
void upSp(int ratio, double* org, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy) {
double* ups;
int width = bmpInfo.biWidth, height = bmpInfo.biHeight;
ups = (double*)calloc(width * height * ratio * ratio, sizeof(double));
// n - tap 보간법 적용
nTapInterpolation(ratio, org, ups, bmpFile, bmpInfo, 4, upy);
free(ups);
}
void nTapInterpolation(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, int n, double* upy) {
// 원본 이미지의 너비와 높이
int originalWidth = bmpInfo.biWidth, originalHeight = bmpInfo.biHeight;
// 업스케일된 이미지의 너비와 높이
int upscaledWidth = originalWidth * ratio, upscaledHeight = originalHeight * ratio;
// 업스케일된 이미지의 바이트 배열 크기와 이미지 크기 계산
int stride2 = (((bmpInfo.biBitCount / 8) * upscaledWidth) + 3) / 4 * 4, size2 = stride2 * upscaledHeight;
// n-TAP 필터 계수 설정 (예: 3-tap 또는 4-tap)
double* coefficients;
if (n == 3) {
double coeffs[3] = { 0.25, 0.5, 0.25 }; // 3-tap 간단 평균 계수
coefficients = coeffs;
}
else if (n == 4) {
double coeffs[4] = { -1.0 / 16, 9.0 / 16, 9.0 / 16, -1.0 / 16 }; // 4-tap 간단 평균 계수
coefficients = coeffs;
}
else {
return;
}
// 업스케일된 이미지의 각 픽셀에 대해
for (int j = 0; j < upscaledHeight; j++) {
for (int i = 0; i < upscaledWidth; i++) {
// 원본 이미지에서 대응되는 좌표 계산
float x = (float)i / ratio;
float y = (float)j / ratio;
double sum = 0;
// n-TAP 필터링
for (int k = 0; k < n; k++) {
int x1 = fmax(0, (int)x - k + 1);
int y1 = fmax(0, (int)y - k + 1);
int x2 = fmin(originalWidth - 1, (int)x + k);
int y2 = fmin(originalHeight - 1, (int)y + k);
sum += org[y1 * originalWidth + x1] * coefficients[k];
sum += org[y2 * originalWidth + x2] * coefficients[k];
}
// 업스케일된 이미지 배열에 값 할당
ups[j * upscaledWidth + i] = sum / 2.0;
upy[j * upscaledWidth + i] = ups[j * upscaledWidth + i];
}
}
}
// Median filter 함수
double median(double* arr, int size) {
// 배열을 정렬
for (int i = 0; i < size - 1; i++) {
for (int j = i + 1; j < size; j++) {
if (arr[i] > arr[j]) {
double tmp = arr[i];
arr[i] = arr[j];
arr[j] = tmp;
}
}
}
// 중간값 반환
if (size % 2 == 0) {
return (arr[size / 2 - 1] + arr[size / 2]) / 2.0;
}
else {
return arr[size / 2];
}
}
// 미디언 필터 적용 함수
void calMedianFilter(double* y3, double* py, int fsize, int height, int width, int pwidth) {
int filterSize = fsize * fsize;
double* window = (double*)malloc(filterSize * sizeof(double));
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
int count = 0;
// 각 픽셀 주변의 값을 window 배열에 저장
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
int xIdx = i + l;
int yIdx = j + k;
// 이미지 경계 처리
if (xIdx >= 0 && xIdx < width && yIdx >= 0 && yIdx < height) {
window[count++] = py[yIdx * pwidth + xIdx];
}
}
}
// 미디언 값을 결과 배열에 저장
y3[j * width + i] = median(window, count);
}
}
free(window);
}
//출력 파일
void makeOutFile(BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, char* fileName, double* y) {
int width = bmpInfo.biWidth, height = bmpInfo.biHeight, size = bmpInfo.biSizeImage, bitCnt = bmpInfo.biBitCount, stride = (((bitCnt / 8) * width) + 3) / 4 * 4;
unsigned char* outputImg = NULL;
outputImg = (unsigned char*)calloc(size, sizeof(unsigned char));
// 값 집어넣어 주기
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
//255 넘지 않게, RGB칸 모두 넣어주기
for (int k = 0; k < 3; k++) outputImg[j * stride + 3 * i + k] = (unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]));
}
}
// 파일 만드는 곳
FILE* outputFile = fopen(fileName, "wb");
fwrite(&bmpFile, sizeof(BITMAPFILEHEADER), 1, outputFile);
fwrite(&bmpInfo, sizeof(BITMAPINFOHEADER), 1, outputFile);
fwrite(outputImg, sizeof(unsigned char), size, outputFile);
fclose(outputFile);
free(outputImg);
}
// 소벨 엣지 멥을 통해 엣지 멥 가져오기
void sobelEdgeMap(double* y, int width, int height, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* edgey) {
//padding
int padding = 2, pwidth = width + padding * 2, pheight = height + padding * 2, psize = pwidth * pheight;
double* py, * xy, * yy, * outy, sumf = 0;
// 값 저장 공간
py = (double*)calloc(pwidth * pheight, sizeof(double));
xy = (double*)calloc(width * height, sizeof(double));
yy = (double*)calloc(width * height, sizeof(double));
outy = (double*)calloc(width * height, sizeof(double));
int fsize = padding * 2 + 1;
// 필터 저장 공간
double* xfilter, * yfilter;
xfilter = (double*)calloc(fsize * fsize, sizeof(double));
yfilter = (double*)calloc(fsize * fsize, sizeof(double));
double xco, yco;
//소벨 필터 만들기
for (int i = 0; i < fsize; i++) {
for (int j = 0; j < fsize; j++) {
xco = (i - padding);
yco = (j - padding);
// 중앙을 기준으로 값 계산
if (xco == 0 && yco == 0) {
xfilter[i * fsize + j] = 0;
yfilter[i * fsize + j] = 0;
}
else {
xfilter[i * fsize + j] = xco / (xco * xco + yco * yco);
yfilter[i * fsize + j] = yco / (xco * xco + yco * yco);
}
}
}
// 패딩 만들기
applyPaddingWithAverage(py, height, width, padding, y);
// 소벨 필터 계산
calFilter(xy, py, fsize, xfilter, height, width, pwidth);
calFilter(yy, py, fsize, yfilter, height, width, pwidth);
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
outy[j * width + i] = sqrt(xy[j * width + i] * xy[j * width + i] + yy[j * width + i] * yy[j * width + i]);
}
}
// 히스토그램 누적 계산을 통한 임계치 구하기
CumulativeHisto(height, width, outy, bmpInfo.biSizeImage, (((bmpInfo.biBitCount / 8) * width) + 3) / 4 * 4, bmpFile, bmpInfo);
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
if (outy[j * width + i] > threshold) edgey[j * width + i] = 254;
}
}
//메모리 해제
free(py);
free(xy);
free(yy);
free(outy);
free(xfilter);
free(yfilter);
}
void bilateralFilter(double* paddedImg, double* outputImg, int width, int height, int pwidth, int pheight, int d, double sigmaColor, double sigmaSpace) {
// 시그마 컬러 : 픽셀값 차이에 대한 민감도 조절
// 시그마 스페이스 : 공간적으로 멀리 떨어진 픽셀들의 영향을 줄인다.
// 필터 반경 설정
int radius = d / 2;
// 시그마 값의 제곱 계산 (색상 및 공간)
double sigmaColor2 = sigmaColor * sigmaColor;
double sigmaSpace2 = sigmaSpace * sigmaSpace;
// 출력 이미지의 각 픽셀에 대해
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
double iFiltered = 0;
double wP = 0;
// 현재 픽셀 값 가져오기
double iValue = paddedImg[(y + radius) * pwidth + (x + radius)];
// 필터링 영역 내의 각 픽셀에 대해
for (int dy = -radius; dy <= radius; dy++) {
for (int dx = -radius; dx <= radius; dx++) {
int ny = y + radius + dy;
int nx = x + radius + dx;
// 이웃 픽셀 값 가져오기
double nValue = paddedImg[ny * pwidth + nx];
// 공간 거리 계산
double spatialDist = dx * dx + dy * dy;
// 색상 거리 계산
double colorDist = (iValue - nValue) * (iValue - nValue);
// 가중치 계산
double weight = exp(-(spatialDist / (2 * sigmaSpace2) + colorDist / (2 * sigmaColor2)));
// 필터링된 값 계산
iFiltered += nValue * weight;
wP += weight;
}
}
// 필터링된 결과를 출력 이미지에 저장
outputImg[y * width + x] = iFiltered / wP;
}
}
}
// 하이브리드 필터 함수
void hybridFilter(double* y3, double* py, double* edgey, int fsize, int height, int width, int pwidth, double* gFilter) {
double* window = (double*)malloc(fsize * fsize * sizeof(double));
double sumf;
// edge에는 미디안, edge가 아닌부분엔 가우시안 필터 적용
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
if (edgey[j * width + i] > 0) {
// Median filter 적용
int count = 0;
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
// 미디안 필터 부분과 똑같다.
int xIdx = i + l;
int yIdx = j + k;
if (xIdx >= 0 && xIdx < width && yIdx >= 0 && yIdx < height) {
window[count++] = py[yIdx * pwidth + xIdx];
}
}
}
y3[j * width + i] = median(window, count);
}
else {
// Gaussian filter 적용
sumf = 0;
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
sumf += gFilter[k * fsize + l] * py[(j + k) * pwidth + (i + l)];
}
}
y3[j * width + i] = sumf;
}
}
}
free(window);
}
void generateGaussianFilter(int n, double sigma, double filter[]) {
double sum = 0.0;
int half = n / 2;
double pi = 3.14159265359;
// Calculate the Gaussian filter
for (int x = -half; x <= half; x++) {
for (int y = -half; y <= half; y++) {
filter[(x + half) * n + (y + half)] = (1 / (2 * pi * sigma * sigma)) * exp(-(x * x + y * y) / (2 * sigma * sigma));
sum += filter[(x + half) * n + (y + half)];
}
}
// Normalize the filter
for (int i = 0; i < n * n; i++) {
filter[i] /= sum;
}
}그래도 이게 제일 준수한 결과를 보였었네요...
노이즈가 있는 사진을 노이즈 제거 및 업샘플링 하는 코드이다.
최대한 주석은 자세하게 작성하였고, 그래도 그나마 결과가 잘 나오는 편이었다.
Edge를 잡았을 때 딱 숫자만 잡아낸 경우이다.
업샘플링 결과인데 사진을 여러장 찍어서 그런지 위 사진을 업샘플링 한 것 같지는 않다.
마지막으로 edge나 히스토그램 사진만 몇개 넣고 종료하겠다.
이건 Data파일에 사진을 저장해 놓으면 알아서 평균 PSNR 구해주는 C 파일입니다.
#define _CRT_SECURE_NO_WARNINGS
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <windows.h>
#include "dirent.h" // Include dirent.h
# define M_PI 3.14159265359
double threshold = 0;
void CumulativeHisto(int height, int width, double* y, int size, int stride, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo);
void applyPaddingWithAverage(double* paddedImg, int height, int width, int padding, double* origImg);
void calFilter(double* y3, double* py, int fsize, double* filter, int height, int width, int pwidth);
double median(double* arr, int size);
void calMedianFilter(double* y3, double* py, int fsize, int height, int width, int pwidth);
void bilateralFilter(double* paddedImg, double* outputImg, int width, int height, int pwidth, int pheight, int d, double sigmaColor, double sigmaSpace);
void getQulity(double* y, double* y3, int width, int height, int bitCnt, double* psnr);
void upSp(int ratio, double* org, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy);
void bilinearUpsampling(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy);
void bilinearUpsample(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy);
void nTapInterpolation(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, int n, double* upy);
void makeOutFile(BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, char* fileName, double* y);
void sobelEdgeMap(double* y, int width, int height, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* edgey);
void hybridFilter(double* y3, double* py, double* edgey, int fsize, int height, int width, int pwidth, double* gFilter);
double cubicInterpolate(double p0, double p1, double p2, double p3, double x);
double bicubicInterpolate(double p[4][4], double x, double y);
void bicubicUpsample(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy);
void generateGaussianFilter(int n, double sigma, double filter[]) {
double sum = 0.0;
int half = n / 2;
// Calculate the Gaussian filter
for (int x = -half; x <= half; x++) {
for (int y = -half; y <= half; y++) {
filter[(x + half) * n + (y + half)] = (1 / (2 * M_PI * sigma * sigma)) * exp(-(x * x + y * y) / (2 * sigma * sigma));
sum += filter[(x + half) * n + (y + half)];
}
}
// Normalize the filter
for (int i = 0; i < n * n; i++) {
filter[i] /= sum;
}
}
void processImagePairs(const char* dirPath) {
DIR* dir;
struct dirent* ent;
FILE* inputFile = NULL, * inputFile2 = NULL;
BITMAPFILEHEADER bmpFile, bmpFile2;
BITMAPINFOHEADER bmpInfo, bmpInfo2;
double totalPsnr = 0;
int count = 0;
if ((dir = opendir(dirPath)) != NULL) {
while ((ent = readdir(dir)) != NULL) {
if (strstr(ent->d_name, "y.bmp") != NULL && !strstr(ent->d_name, "yy.bmp")) {
// Construct the corresponding "yy" file name
char filePath1[256], filePath2[256];
snprintf(filePath1, sizeof(filePath1), "%s/%s", dirPath, ent->d_name);
// Replace "y.bmp" with "yy.bmp"
strcpy(filePath2, filePath1);
strcpy(filePath2 + strlen(filePath2) - 5, "yy.bmp");
inputFile = fopen(filePath1, "rb");
inputFile2 = fopen(filePath2, "rb");
if (inputFile == NULL || inputFile2 == NULL) {
printf("Unable to open file %s or %s\n", filePath1, filePath2);
if (inputFile) fclose(inputFile);
if (inputFile2) fclose(inputFile2);
continue;
}
fread(&bmpFile, sizeof(BITMAPFILEHEADER), 1, inputFile);
fread(&bmpInfo, sizeof(BITMAPINFOHEADER), 1, inputFile);
fread(&bmpFile2, sizeof(BITMAPFILEHEADER), 1, inputFile2);
fread(&bmpInfo2, sizeof(BITMAPINFOHEADER), 1, inputFile2);
int width = bmpInfo.biWidth, height = bmpInfo.biHeight, size = bmpInfo.biSizeImage, bitCnt = bmpInfo.biBitCount, stride = (((bitCnt / 8) * width) + 3) / 4 * 4;
int width2 = bmpInfo2.biWidth, height2 = bmpInfo2.biHeight, size2 = bmpInfo2.biSizeImage, bitCnt2 = bmpInfo2.biBitCount, stride2 = (((bitCnt2 / 8) * width2) + 3) / 4 * 4;
double* y = (double*)calloc(width * height, sizeof(double));
double* y2 = (double*)calloc(width * height, sizeof(double));
double mse = 0, psnr = 0, * edgey;
edgey = (double*)calloc(width2 * height2, sizeof(double));
unsigned char* inputImg = (unsigned char*)calloc(size, sizeof(unsigned char));
unsigned char* inputImg2 = (unsigned char*)calloc(size, sizeof(unsigned char));
fread(inputImg, sizeof(unsigned char), size, inputFile);
fread(inputImg2, sizeof(unsigned char), size, inputFile2);
printf("Processing files: %s and %s\n", filePath1, filePath2);
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
y[j * width + i] = inputImg[j * stride + 3 * i];
}
}
for (int j = 0; j < height2; j++) {
for (int i = 0; i < width2; i++) {
y2[j * width2 + i] = inputImg2[j * stride2 + 3 * i];
}
}
// Algorithm
// sampling
int ratio = 4;
double* upy;
upy = (double*)calloc(width * height * ratio * ratio, sizeof(double));
//padding
int padding = 2, pwidth = width2 + padding * 2, pheight = height2 + padding * 2, psize = pwidth * pheight;
double* py;
py = (double*)calloc(pwidth * pheight, sizeof(double));
applyPaddingWithAverage(py, height2, width2, padding, y2);
// filter
int fsize = padding * 2 + 1;
// Gussian Filter
double G3filter[] = { 1.0 / 16,2.0 / 16,1.0 / 16,2.0 / 16,4.0 / 16,2.0 / 16,1.0 / 16,2.0 / 16,1.0 / 16 };
double G5filter[] = { 1 / 273.0, 4 / 273.0, 7 / 273.0, 4 / 273.0, 1 / 273.0, 4 / 273.0, 16 / 273.0, 26 / 273.0, 16 / 273.0, 4 / 273.0, 7 / 273.0, 26 / 273.0, 41 / 273.0, 26 / 273.0, 7 / 273.0, 4 / 273.0, 16 / 273.0, 26 / 273.0, 16 / 273.0, 4 / 273.0,1 / 273.0, 4 / 273.0, 7 / 273.0, 4 / 273.0, 1 / 273.0 };
double* y3;
y3 = (double*)calloc(width2 * height2, sizeof(double));
//calFilter(y3,py, fsize, G3filter,height2,width2,pwidth);
//calMedianFilter( y3,py, fsize, height2, width2, pwidth);
bilateralFilter(py, y3, width2, height2, pwidth, pheight, fsize, 50.0, 10); //25.29
//sobelEdgeMap(y3, width2, height2, bmpFile2, bmpInfo2, edgey);
//applyPaddingWithAverage(py, height2, width2, padding, y3);
//bilateralFilter(py, y3, width2, height2, pwidth, pheight, fsize, 50.0, 10);
//calMedianFilter(y3, py, fsize, height2, width2, pwidth);
//hybridFilter(y3, py, edgey, fsize, height2, width2, pwidth, G5filter);
//get Quality
upSp(ratio, y3, bmpFile2, bmpInfo2, upy);
double* edgey2;
edgey2 = (double*)calloc(width * height, sizeof(double));
int padding2 = 2, pwidth2 = width + padding2 * 2, pheight2 = height + padding2 * 2, psize2 = pwidth2 * pheight2;
double* py2;
py2 = (double*)calloc(pwidth2 * pheight2, sizeof(double));
int fsize2 = padding2 * 2 + 1;
applyPaddingWithAverage(py2, height, width, padding2, upy);
bilateralFilter(py2, upy, width, height, pwidth2, pheight2, fsize2, 60.0, 10.0);//192.72
applyPaddingWithAverage(py2, height, width, padding2, upy);
sobelEdgeMap(upy, width, height, bmpFile, bmpInfo, edgey2);
hybridFilter(upy, py2, edgey2, fsize2, height, width, pwidth2, G5filter);
//applyPaddingWithAverage(py2, height, width, padding2, upy);
//sobelEdgeMap(upy, width, height, bmpFile, bmpInfo, edgey2);
//hybridFilter(upy, py2, edgey2, fsize2, height, width, pwidth2, G5filter);
int padding3 = 1, pwidth3 = width + padding3 * 2, pheight3 = height + padding3 * 2, psize3 = pwidth3 * pheight3;
double* py3;
py3 = (double*)calloc(pwidth3 * pheight3, sizeof(double));
int fsize3 = padding3 * 2 + 1;
double* filter;
filter = (double*)calloc(fsize3 * fsize3, sizeof(double));
for (int i = 0; i < fsize3 * fsize3; i++) filter[i] = 1.0 / (fsize3 * fsize3);
double *gfilter;
gfilter = (double*)calloc(fsize3 * fsize3, sizeof(double));
generateGaussianFilter(fsize3, fsize3 / 6.0, gfilter);
applyPaddingWithAverage(py3, height, width, padding3, upy);
sobelEdgeMap(upy, width, height, bmpFile, bmpInfo, edgey2);
hybridFilter(upy, py3, edgey2, fsize3, height, width, pwidth3, gfilter);
applyPaddingWithAverage(py2, height, width, padding2, upy);
calMedianFilter(upy, py2, fsize2, height, width, pwidth2);
getQulity(y, upy, width, height, bitCnt, &psnr);
totalPsnr += psnr;
count++;
free(inputImg);
free(inputImg2);
free(y);
free(y2);
free(py);
free(y3);
free(edgey);
free(upy);
free(edgey2);
free(py2);
fclose(inputFile);
fclose(inputFile2);
}
}
closedir(dir);
}
else {
perror("Could not open directory");
}
if (count > 0) {
printf("Average PSNR: %.2lf\n", totalPsnr / count);
}
else {
printf("No valid image pairs found.\n");
}
}
int main() {
processImagePairs("data");
return 0;
}
void getQulity(double* y, double* y3, int width, int height, int bitCnt, double* psnr) {
double mse = 0, max;
unsigned char org, out;
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
org = (unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]));
out = (unsigned char)(y3[j * width + i] > 255 ? 255 : (y3[j * width + i] < 0 ? 0 : y3[j * width + i]));
mse += (double)((org - out) * (org - out));
}
}
mse /= (height * width);
max = pow(2, bitCnt / 3) - 1;
*psnr = mse != 0.0 ? 10.0 * log10(max * max / mse) : 99.99;
printf("mse = %.2lf\npsnr = %.2lf\n", mse, *psnr);
}
// 나머지 함수는 그대로 유지됩니다.
void calFilter(double* y3, double* py, int fsize, double* filter, int height, int width, int pwidth) {
double sumf = 0;
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
sumf = 0;
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
sumf += filter[k * fsize + l] * py[(j + k) * pwidth + (i + l)];
}
}
y3[j * width + i] = sumf;
}
}
}
void CumulativeHisto(int height, int width, double* y, int size, int stride, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo) {
double* histogram;
histogram = (double*)calloc(256, sizeof(double));
int width2 = 512, height2 = 512, stride2 = (((bmpInfo.biBitCount / 8) * width2) + 3) / 4 * 4, size2 = stride2 * height2;
// get Histogram
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
histogram[(unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]))]++;
}
}
for (int i = 0; i < 255; i++) {
histogram[i + 1] += histogram[i];
}
for (int i = 0; i < 255; i++) {
if (histogram[i] > height * width * 0.985) {
threshold = i;
break;
}
if (i == 254) threshold = 254;
}
// out Histogram
double histoMax = 0;
for (int i = 0; i < 256; i++) {
if (histogram[i] > histoMax) histoMax = histogram[i];
}
unsigned char* outHisto, * HE;
outHisto = (unsigned char*)calloc(size2, sizeof(unsigned char));
HE = (unsigned char*)calloc(size, sizeof(unsigned char));
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
HE[j * stride + 3 * i + 0] = (unsigned char)(255 * histogram[(unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]))] / (width * height));
HE[j * stride + 3 * i + 1] = (unsigned char)(255 * histogram[(unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]))] / (width * height));
HE[j * stride + 3 * i + 2] = (unsigned char)(255 * histogram[(unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]))] / (width * height));
}
}
for (int i = 0; i < 256; i++)histogram[i] = histogram[i] / histoMax * 450;
for (int j = 0; j < height2; j++) {
for (int i = 0; i < width2; i++) {
if (histogram[i / 2] > j) continue;
outHisto[j * stride2 + 3 * i + 0] = 255;
outHisto[j * stride2 + 3 * i + 1] = 255;
outHisto[j * stride2 + 3 * i + 2] = 255;
}
}
FILE* outputHisto = fopen("OutCHisto.bmp", "wb");
bmpInfo.biWidth = 512;
bmpInfo.biHeight = 512;
bmpInfo.biSizeImage = 512 * 3 * 512;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + 512 * 3 * 512;
fwrite(&bmpFile, sizeof(BITMAPFILEHEADER), 1, outputHisto);
fwrite(&bmpInfo, sizeof(BITMAPINFOHEADER), 1, outputHisto);
fwrite(outHisto, sizeof(unsigned char), size2, outputHisto);
FILE* outputHE = fopen("OutHE.bmp", "wb");
bmpInfo.biWidth = width;
bmpInfo.biHeight = height;
bmpInfo.biSizeImage = size;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + size;
fwrite(&bmpFile, sizeof(BITMAPFILEHEADER), 1, outputHE);
fwrite(&bmpInfo, sizeof(BITMAPINFOHEADER), 1, outputHE);
fwrite(HE, sizeof(unsigned char), size, outputHE);
fclose(outputHE);
free(HE);
free(histogram);
fclose(outputHisto);
free(outHisto);
}
void applyPaddingWithAverage(double* paddedImg, int height, int width, int padding, double* origImg) {
int pheight = height + 2 * padding;
int pwidth = width + 2 * padding;
// 원본 이미지 데이터를 중앙에 배치
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
paddedImg[(j + padding) * pwidth + (i + padding)] = origImg[j * width + i];
}
}
// 상하좌우 패딩에 평균 값 적용
for (int j = 0; j < pheight; j++) {
for (int i = 0; i < pwidth; i++) {
// 좌측 패딩 처리
if (i < padding) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = j + k; // 현재 행에서 상하로 2개씩
if (idx >= padding && idx < height + padding) { // 범위 체크
sum += paddedImg[idx * pwidth + padding];
count++;
}
}
paddedImg[j * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
// 우측 패딩 처리
if (i >= width + padding) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = j + k; // 현재 행에서 상하로 2개씩
if (idx >= padding && idx < height + padding) { // 범위 체크
sum += paddedImg[idx * pwidth + (width + padding - 1)];
count++;
}
}
paddedImg[j * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
// 상단 패딩 처리
for (int p = 0; p < padding; p++) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = i + k; // 현재 컬럼에서 좌우로 2개씩
if (idx >= padding && idx < width + padding) { // 범위 체크
sum += paddedImg[padding * pwidth + idx];
count++;
}
}
paddedImg[p * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
// 하단 패딩 처리
for (int p = 0; p < padding; p++) {
double sum = 0;
int count = 0;
for (int k = -2; k <= 2; k++) {
int idx = i + k; // 현재 컬럼에서 좌우로 2개씩
if (idx >= padding && idx < width + padding) { // 범위 체크
sum += paddedImg[(height + padding - 1) * pwidth + idx];
count++;
}
}
paddedImg[(height + padding + p) * pwidth + i] = count > 0 ? sum / count : 0; // 평균 값 적용
}
}
}
}
void upSp(int ratio, double* org, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy) {
double* ups;
int width = bmpInfo.biWidth, height = bmpInfo.biHeight;
//ups = (double*)calloc(width * height * ratio/2 * ratio/2, sizeof(double));
ups = (double*)calloc(width * height * ratio * ratio, sizeof(double));
//upsampling(ratio/2, org, ups, bmpFile, bmpInfo,upy); // 25.23
//bilinearUpsample(ratio, org, ups, bmpFile, bmpInfo, upy); // 25.22
//bicubicUpsample(ratio, org, ups, bmpFile, bmpInfo, upy);
//bilinearUpsampling(ratio/2, org, ups, bmpFile, bmpInfo, upy);//25.22
nTapInterpolation(ratio, org, ups, bmpFile, bmpInfo, 4, upy); // 25.19
double* ups2;
bmpInfo.biHeight = bmpInfo.biHeight * ratio / 2;
bmpInfo.biWidth = bmpInfo.biWidth * ratio / 2;
ups2 = (double*)calloc(width * height * ratio * ratio, sizeof(double));
//upsampling(ratio / 2, ups, ups2, bmpFile, bmpInfo, upy);
//bilinearUpsampling(ratio / 2, ups, ups2, bmpFile, bmpInfo, upy); //24.90
free(ups);
}
void bilinearUpsampling(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy) {
int originalWidth = bmpInfo.biWidth, originalHeight = bmpInfo.biHeight;
int upscaledWidth = bmpInfo.biWidth * ratio, upscaledHeight = bmpInfo.biHeight * ratio;
int stride2 = (((bmpInfo.biBitCount / 8) * upscaledWidth) + 3) / 4 * 4, size2 = stride2 * upscaledHeight;
for (int y = 0; y < upscaledHeight; y++) {
for (int x = 0; x < upscaledWidth; x++) {
float gx = ((float)x / (float)upscaledWidth) * (float)(originalWidth - 1);
float gy = ((float)y / (float)upscaledHeight) * (float)(originalHeight - 1);
int gxi = (int)gx;
int gyi = (int)gy;
double top = (gx - gxi) * org[gyi * originalWidth + (gxi + 1)] + (gxi + 1 - gx) * org[gyi * originalWidth + gxi];
double bottom = (gx - gxi) * org[(gyi + 1) * originalWidth + (gxi + 1)] + (gxi + 1 - gx) * org[(gyi + 1) * originalWidth + gxi];
ups[y * upscaledWidth + x] = (gy - gyi) * bottom + (gyi + 1 - gy) * top;
upy[y * upscaledWidth + x] = ups[y * upscaledWidth + x];
}
}
bmpInfo.biWidth = upscaledWidth;
bmpInfo.biHeight = upscaledHeight;
bmpInfo.biSizeImage = size2;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + size2;
makeOutFile(bmpFile, bmpInfo, "OutUPS.bmp", ups);
}
void bilinearUpsample(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy) {
// 원본 이미지의 너비와 높이
int originalWidth = bmpInfo.biWidth, originalHeight = bmpInfo.biHeight;
// 업스케일된 이미지의 너비와 높이
int upscaledWidth = originalWidth * ratio, upscaledHeight = originalHeight * ratio;
// 업스케일된 이미지의 바이트 배열 크기와 이미지 크기 계산
int stride2 = (((bmpInfo.biBitCount / 8) * upscaledWidth) + 3) / 4 * 4, size2 = stride2 * upscaledHeight;
// 업스케일된 이미지의 각 픽셀에 대해
for (int y = 0; y < upscaledHeight; y++) {
for (int x = 0; x < upscaledWidth; x++) {
// 원본 이미지에서 대응되는 좌표 계산
float gx = ((float)x / (float)upscaledWidth) * (float)(originalWidth - 1);
float gy = ((float)y / (float)upscaledHeight) * (float)(originalHeight - 1);
int gxi = (int)gx;
int gyi = (int)gy;
// 4개의 인접한 픽셀값 가져오기 (원본 이미지 경계 검사 포함)
double c00 = org[gyi * originalWidth + gxi];
double c10 = org[gyi * originalWidth + (gxi + 1 < originalWidth ? gxi + 1 : gxi)];
double c01 = org[(gyi + 1 < originalHeight ? gyi + 1 : gyi) * originalWidth + gxi];
double c11 = org[(gyi + 1 < originalHeight ? gyi + 1 : gyi) * originalWidth + (gxi + 1 < originalWidth ? gxi + 1 : gxi)];
// 두 좌표 사이의 거리 계산
double dx = gx - gxi;
double dy = gy - gyi;
// 두 방향으로 보간
double c0 = c00 * (1 - dx) + c10 * dx;
double c1 = c01 * (1 - dx) + c11 * dx;
double c = c0 * (1 - dy) + c1 * dy;
// 업스케일된 이미지 배열에 값 할당
ups[y * upscaledWidth + x] = c;
upy[y * upscaledWidth + x] = ups[y * upscaledWidth + x];
}
}
// 비트맵 헤더 정보 업데이트
bmpInfo.biWidth = upscaledWidth;
bmpInfo.biHeight = upscaledHeight;
bmpInfo.biSizeImage = size2;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + size2;
// 결과 파일 저장
makeOutFile(bmpFile, bmpInfo, "OutBiUPS.bmp", ups);
}
void nTapInterpolation(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, int n, double* upy) {
// 원본 이미지의 너비와 높이
int originalWidth = bmpInfo.biWidth, originalHeight = bmpInfo.biHeight;
// 업스케일된 이미지의 너비와 높이
int upscaledWidth = originalWidth * ratio, upscaledHeight = originalHeight * ratio;
// 업스케일된 이미지의 바이트 배열 크기와 이미지 크기 계산
int stride2 = (((bmpInfo.biBitCount / 8) * upscaledWidth) + 3) / 4 * 4, size2 = stride2 * upscaledHeight;
// n-TAP 필터 계수 설정 (예: 3-tap 또는 4-tap)
double* coefficients;
if (n == 3) {
double coeffs[3] = { 0.25, 0.5, 0.25 }; // 3-tap 간단 평균 계수
coefficients = coeffs;
}
else if (n == 4) {
double coeffs[4] = { -1.0 / 16, 9.0 / 16, 9.0 / 16, -1.0 / 16 }; // 4-tap 간단 평균 계수
coefficients = coeffs;
}
else {
printf("지원되지 않는 n-tap 값: %d\n", n);
return;
}
// 업스케일된 이미지의 각 픽셀에 대해
for (int j = 0; j < upscaledHeight; j++) {
for (int i = 0; i < upscaledWidth; i++) {
// 원본 이미지에서 대응되는 좌표 계산
float x = (float)i / ratio;
float y = (float)j / ratio;
double sum = 0;
// n-TAP 필터링
for (int k = 0; k < n; k++) {
int x1 = fmax(0, (int)x - k + 1);
int y1 = fmax(0, (int)y - k + 1);
int x2 = fmin(originalWidth - 1, (int)x + k);
int y2 = fmin(originalHeight - 1, (int)y + k);
sum += org[y1 * originalWidth + x1] * coefficients[k];
sum += org[y2 * originalWidth + x2] * coefficients[k];
}
// 업스케일된 이미지 배열에 값 할당
ups[j * upscaledWidth + i] = sum / 2;
upy[j * upscaledWidth + i] = ups[j * upscaledWidth + i];
}
}
// 비트맵 헤더 정보 업데이트
bmpInfo.biWidth = upscaledWidth;
bmpInfo.biHeight = upscaledHeight;
bmpInfo.biSizeImage = size2;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + size2;
// 결과 파일 저장
makeOutFile(bmpFile, bmpInfo, "OutUPS.bmp", ups);
}
double cubicInterpolate(double p0, double p1, double p2, double p3, double x) {
return p1 + 0.5 * x * (p2 - p0 + x * (2.0 * p0 - 5.0 * p1 + 4.0 * p2 - p3 + x * (3.0 * (p1 - p2) + p3 - p0)));
}
double bicubicInterpolate(double p[4][4], double x, double y) {
double arr[4];
arr[0] = cubicInterpolate(p[0][0], p[0][1], p[0][2], p[0][3], x);
arr[1] = cubicInterpolate(p[1][0], p[1][1], p[1][2], p[1][3], x);
arr[2] = cubicInterpolate(p[2][0], p[2][1], p[2][2], p[2][3], x);
arr[3] = cubicInterpolate(p[3][0], p[3][1], p[3][2], p[3][3], x);
return cubicInterpolate(arr[0], arr[1], arr[2], arr[3], y);
}
void bicubicUpsample(int ratio, double* org, double* ups, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* upy) {
// 원본 이미지의 너비와 높이
int originalWidth = bmpInfo.biWidth, originalHeight = bmpInfo.biHeight;
// 업스케일된 이미지의 너비와 높이
int upscaledWidth = originalWidth * ratio, upscaledHeight = originalHeight * ratio;
// 업스케일된 이미지의 바이트 배열 크기와 이미지 크기 계산
int stride2 = (((bmpInfo.biBitCount / 8) * upscaledWidth) + 3) / 4 * 4, size2 = stride2 * upscaledHeight;
for (int j = 0; j < upscaledHeight; j++) {
for (int i = 0; i < upscaledWidth; i++) {
// 원본 이미지에서 대응되는 좌표 계산
float gx = ((float)i / (float)upscaledWidth) * (float)(originalWidth - 1);
float gy = ((float)j / (float)upscaledHeight) * (float)(originalHeight - 1);
int gxi = (int)gx;
int gyi = (int)gy;
// 4x4 블록에서 픽셀 값을 가져옴
double p[4][4];
for (int m = -1; m <= 2; m++) {
for (int n = -1; n <= 2; n++) {
int x = min(max(gxi + n, 0), originalWidth - 1);
int y = min(max(gyi + m, 0), originalHeight - 1);
p[m + 1][n + 1] = org[y * originalWidth + x];
}
}
// 보간 계산
double dx = gx - gxi;
double dy = gy - gyi;
double value = bicubicInterpolate(p, dx, dy);
// 업스케일된 이미지 배열에 값 할당
ups[j * upscaledWidth + i] = value;
upy[j * upscaledWidth + i] = ups[j * upscaledWidth + i];
}
}
// 비트맵 헤더 정보 업데이트
bmpInfo.biWidth = upscaledWidth;
bmpInfo.biHeight = upscaledHeight;
bmpInfo.biSizeImage = size2;
bmpFile.bfSize = sizeof(BITMAPFILEHEADER) + sizeof(BITMAPINFOHEADER) + size2;
// 결과 파일 저장
makeOutFile(bmpFile, bmpInfo, "OutBicubicUPS.bmp", ups);
}
// Median filter 함수
double median(double* arr, int size) {
// 배열을 정렬
for (int i = 0; i < size - 1; i++) {
for (int j = i + 1; j < size; j++) {
if (arr[i] > arr[j]) {
double tmp = arr[i];
arr[i] = arr[j];
arr[j] = tmp;
}
}
}
// 중간값 반환
if (size % 2 == 0) {
return (arr[size / 2 - 1] + arr[size / 2]) / 2.0;
}
else {
return arr[size / 2];
}
}
// 미디언 필터 적용 함수
void calMedianFilter(double* y3, double* py, int fsize, int height, int width, int pwidth) {
int filterSize = fsize * fsize;
double* window = (double*)malloc(filterSize * sizeof(double));
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
int count = 0;
// 각 픽셀 주변의 값을 window 배열에 저장
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
int xIdx = i + l;
int yIdx = j + k;
// 이미지 경계 처리
if (xIdx >= 0 && xIdx < width && yIdx >= 0 && yIdx < height) {
window[count++] = py[yIdx * pwidth + xIdx];
}
}
}
// 미디언 값을 결과 배열에 저장
y3[j * width + i] = median(window, count);
}
}
free(window);
}
void makeOutFile(BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, char* fileName, double* y) {
int width = bmpInfo.biWidth, height = bmpInfo.biHeight, size = bmpInfo.biSizeImage, bitCnt = bmpInfo.biBitCount, stride = (((bitCnt / 8) * width) + 3) / 4 * 4;
unsigned char* outputImg = NULL;
outputImg = (unsigned char*)calloc(size, sizeof(unsigned char));
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
for (int k = 0; k < 3; k++) outputImg[j * stride + 3 * i + k] = (unsigned char)(y[j * width + i] > 255 ? 255 : (y[j * width + i] < 0 ? 0 : y[j * width + i]));
}
}
FILE* outputFile = fopen(fileName, "wb");
fwrite(&bmpFile, sizeof(BITMAPFILEHEADER), 1, outputFile);
fwrite(&bmpInfo, sizeof(BITMAPINFOHEADER), 1, outputFile);
fwrite(outputImg, sizeof(unsigned char), size, outputFile);
fclose(outputFile);
free(outputImg);
}
void sobelEdgeMap(double* y, int width, int height, BITMAPFILEHEADER bmpFile, BITMAPINFOHEADER bmpInfo, double* edgey) {
//padding
int padding = 2, pwidth = width + padding * 2, pheight = height + padding * 2, psize = pwidth * pheight;
double* py, * xy, * yy, * outy, sumf = 0;
py = (double*)calloc(pwidth * pheight, sizeof(double));
xy = (double*)calloc(width * height, sizeof(double));
yy = (double*)calloc(width * height, sizeof(double));
outy = (double*)calloc(width * height, sizeof(double));
int fsize = padding * 2 + 1;
double* xfilter, * yfilter;
xfilter = (double*)calloc(fsize * fsize, sizeof(double));
yfilter = (double*)calloc(fsize * fsize, sizeof(double));
double xco, yco;
for (int i = 0; i < fsize; i++) {
for (int j = 0; j < fsize; j++) {
xco = (i - padding);
yco = (j - padding);
if (xco == 0 && yco == 0) {
xfilter[i * fsize + j] = 0;
yfilter[i * fsize + j] = 0;
}
else {
xfilter[i * fsize + j] = xco / (xco * xco + yco * yco);
yfilter[i * fsize + j] = yco / (xco * xco + yco * yco);
}
}
}
applyPaddingWithAverage(py, height, width, padding, y);
calFilter(xy, py, fsize, xfilter, height, width, pwidth);
calFilter(yy, py, fsize, yfilter, height, width, pwidth);
makeOutFile(bmpFile, bmpInfo, "xEdgeOut.bmp", xy);
makeOutFile(bmpFile, bmpInfo, "yEdgeOut.bmp", yy);
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
outy[j * width + i] = sqrt(xy[j * width + i] * xy[j * width + i] + yy[j * width + i] * yy[j * width + i]);
}
}
CumulativeHisto(height, width, outy, bmpInfo.biSizeImage, (((bmpInfo.biBitCount / 8) * width) + 3) / 4 * 4, bmpFile, bmpInfo);
//printf("threshold= %f\n", threshold);
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
if (outy[j * width + i] > threshold) edgey[j * width + i] = 254;
}
}
makeOutFile(bmpFile, bmpInfo, "data/EdgeOut.bmp", outy);
makeOutFile(bmpFile, bmpInfo, "data/thresholdEdge.bmp", edgey);
free(py);
free(xy);
free(yy);
free(outy);
}
void bilateralFilter(double* paddedImg, double* outputImg, int width, int height, int pwidth, int pheight, int d, double sigmaColor, double sigmaSpace) {
// 시그마 컬러 : 픽셀값 차이에 대한 민감도 조절
// 시그마 스페이스 : 공간적으로 멀리 떨어진 픽셀들의 영향을 줄인다.
// 필터 반경 설정
int radius = d / 2;
// 시그마 값의 제곱 계산 (색상 및 공간)
double sigmaColor2 = sigmaColor * sigmaColor;
double sigmaSpace2 = sigmaSpace * sigmaSpace;
// 출력 이미지의 각 픽셀에 대해
for (int y = 0; y < height; y++) {
for (int x = 0; x < width; x++) {
double iFiltered = 0;
double wP = 0;
// 현재 픽셀 값 가져오기
double iValue = paddedImg[(y + radius) * pwidth + (x + radius)];
// 필터링 영역 내의 각 픽셀에 대해
for (int dy = -radius; dy <= radius; dy++) {
for (int dx = -radius; dx <= radius; dx++) {
int ny = y + radius + dy;
int nx = x + radius + dx;
// 이웃 픽셀 값 가져오기
double nValue = paddedImg[ny * pwidth + nx];
// 공간 거리 계산
double spatialDist = dx * dx + dy * dy;
// 색상 거리 계산
double colorDist = (iValue - nValue) * (iValue - nValue);
// 가중치 계산
double weight = exp(-(spatialDist / (2 * sigmaSpace2) + colorDist / (2 * sigmaColor2)));
// 필터링된 값 계산
iFiltered += nValue * weight;
wP += weight;
}
}
// 필터링된 결과를 출력 이미지에 저장
outputImg[y * width + x] = iFiltered / wP;
}
}
}
// 하이브리드 필터 함수
void hybridFilter(double* y3, double* py, double* edgey, int fsize, int height, int width, int pwidth, double* gFilter) {
double* window = (double*)malloc(fsize * fsize * sizeof(double));
double sumf;
for (int j = 0; j < height; j++) {
for (int i = 0; i < width; i++) {
if (edgey[j * width + i] > 0) {
// Median filter
int count = 0;
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
int xIdx = i + l;
int yIdx = j + k;
if (xIdx >= 0 && xIdx < width && yIdx >= 0 && yIdx < height) {
window[count++] = py[yIdx * pwidth + xIdx];
}
}
}
y3[j * width + i] = median(window, count);
}
else {
// Gaussian filter
sumf = 0;
for (int k = 0; k < fsize; k++) {
for (int l = 0; l < fsize; l++) {
sumf += gFilter[k * fsize + l] * py[(j + k) * pwidth + (i + l)];
}
}
y3[j * width + i] = sumf;
}
}
}
free(window);
}예비군 다녀오자마자 기절했다 깨어나니 이시간이라... 일단 여기까지..
728x90
'언어 > C' 카테고리의 다른 글
| C언어 영상처리 과제 - compression, Quantization (0) | 2024.06.05 |
|---|---|
| C언어 영상처리 - DPCM (0) | 2024.05.23 |
| C언어 영상처리 - data 파일에 넣은 사진 psnr 자동 계산하기 (0) | 2024.05.16 |
| C언어 영상처리 - 노이즈 제거, 업스케일링 기록용 (1) | 2024.05.16 |
| C 언어 영상처리 - Sobel Filter, 윤곽선 검출, Edge Detection (0) | 2024.05.07 |