experiment one:
Write RGB to YUV program, focus on function definition, initialization and calling of some lookup tables, buffer allocation. Convert the obtained RGB file to YUV file and watch it with YUV Viewer player to verify whether it is correct.
Write a program to convert YUV to RGB. Use this program to convert the given experimental data into RGB files. And with the original
RGB files are compared, and if there are errors, analyze where the errors come from.
I wrote a simple version before4:4:4ofRGBFile and4:4:4ofYUVFor file conversion, the link is as follows. This version of the article derives the conversion formula and the realization of sampling based on the previous article. If you only focus on two types of conversion, you can go directly to the non-sampling version to view the code and ideas.
Data compression (four)-color space conversion (non-sampling version)
Article Directory
- (1) Conversion formula and file storage format of YUV and RGB
- 1.1 Derivation of conversion formula
- 1.1.1 Conversion formula of analog signal
- 1.1.2 Conversion formula of digital signal
- Normalized
- Code level distribution after luminance signal quantization
- Code level distribution after color difference signal quantization
- Code level digital expression
- 1.2 File storage format
- (Two) the command line parameters of the main function
- 2.1 Representation
- 2.2 How to use
- (3) Initial realization of color space conversion (no sampling) code
- main.cpp
- rgb2yuv.h
- rgb2yuv.cpp
- yuv2rgb.h
- yuv2rgb.cpp
- Experimental results
- (4) Analysis of the causes of experimental errors and code modification
- Error correction
- Experimental results
- (5) Optimize the code (using the lookup table method)
- main.cpp
- yuvrgb.h
- yuvrgb.cpp
- Experimental results
- (6) Color space conversion (sampling) code implementation
- main.cpp
- yuvrgb.h
- yuvrgb.cpp
- Experimental results
- (7) Analysis of experimental error
R=(298×Y+411×V−57376)>>8G=(298×Y−101×U−211×V+35168)>>8B=(298×Y+519×U−71200)>>8Y=(66×R+129×G+25×B)>>8+16U=(−38×R−74×G+112×B)>>8+128V=(112×R−94×G−18×B)>>8+128
The above formula is already quantified.
1.1 Derivation of conversion formula
1.1.1 Conversion formula of analog signal
According to the knowledge of TV principles, we can know the following formula:
Y=0.2990×R+0.5870×G+0.1140×BU=B−Y=−0.2990×R−0.5870×G+0.8860×BV=R−Y=0.7010×R−0.5870×G−0.1140×B
1.1.2 Conversion formula of digital signal
Normalized
In order to facilitate processing, the analog signal needs to be normalized when it is converted into a digital signal, so that the dynamic range of the color difference signal is controlled within−0.5∼0.5between.
After returning to one sentence, the formula is obtained:
U′=0.564×(B−Y)=0.564×(−0.2990×R−0.5870×G+0.8860×B)V′=0.713×(R−Y)=0.713×(0.7010×R−0.5870×G−0.1140×B)
Due to the aboveU′、V′The value range of±350mvBetween, so you need to introduce+350mv bias. The formula becomes:
U′′=0.564×(B−Y)+0.35V′′=0.713×(R−Y)+0.35
Code level distribution after luminance signal quantization
The code level distribution after the quantization of the luminance signal is shown in the figure (source "Digital TV Broadcasting Principles and Applications" P36):
It can be seen that the peak white level of the brightness signal corresponds to the code level 235, and the blanking level corresponds to the code level 16. In order to prevent overload caused by signal changes, 20 levels at the upper end and 16 levels at the lower end are used as protection bands for the signal to exceed the dynamic range. The code levels 0 and 255 are protection levels, which are not allowed to appear in the video data stream, and the code words 00 and FF are used to transmit synchronization information. The dynamic range of the luminance signal occupies a total of 220 quantization levels.
Code level distribution after color difference signal quantization
The code level distribution of the color difference signal is shown in the figure, the source of the picture is "Digital TV Broadcasting Principles and Applications" P37):
Color difference signalCbwithCrThe peak white level of the peak corresponds to the code level 240, and the 0 level corresponds to the code level 16. In order to prevent overload caused by signal changes, 15 levels at the upper end and 16 levels at the lower end are used as the protection band for the signal to exceed the dynamic range. The dynamic range of the color difference signal occupies a total of 225 quantization levels.
Code level digital expression
After the quantization of the luminance signal and the color difference signal, the nearest integer is taken as the code level value. The digitized expression is:
DY=INT[(219Y+16)×2n−8]DCB=INT[(224Cb+128)×2n−8]DCR=INT[(224Cr+128)×2n−8]
Push the previousY,U′′,V′′Bringing into the above formula, you can get:
Y=(66×R+129×G+25×B)>>8+16U=(−38×R−74×G+112×B)>>8+128V=(112×R−94×G−18×B)>>8+128
Write the expression in the form of a matrix:
⎣⎡YUV⎦⎤=2561⎣⎡66−38112129−74−9425112−18⎦⎤⎣⎡RGB⎦⎤+⎣⎡16128128⎦⎤
order matrix⎣⎡66−38112129−74−9425112−18⎦⎤MatrixA, The original formula can be transformed into:⎣⎡RGB⎦⎤=256A−1⎣⎡Y−16U−128V−128⎦⎤
Because ofA−1Is too small, so firstAEach element in is reduced to the original256×2561getA1. Then the above formula is converted to:⎣⎡RGB⎦⎤=A1−1⎣⎡Y−16G−128B−128⎦⎤÷256
Rounding the coefficient to the whole, the following formula can be obtained:
R=(298×Y+411×V−57376)>>8G=(298×Y−101×U−211×V+35168)>>8B=(298×Y+519×U−71200)>>8
1.2 File storage format
Check the information,RGBThe file storage format isBGRBGRBGR⋯,YUVThe file storage format is first save allY, Save allUAnd finally save allV。
2.1 Representation
One programmain()The function can contain two parameters:
- The first parameter isintTypes of;
- The second parameter is a string array;
Usually, the first parameter is namedargc, The second parameter isargv. Since the declaration of the string array in the function header can have two forms, somain()There are also two ways to write functions.
-
main()functionWriting one:
int main(int argc, char** argv){ return 0;}
-
main()functionWriting two:
int main(int argc, char* argv[]){ return 0;}
2.2 How to use
-
The meaning of the parameters:
int argc: Represents the number of strings.argc = 1 + the number of strings entered by the user, The value of argc is calculated automatically by the operating system, and the programmer does not need to assign it.
char argv[]*: It stores multiple strings, the form of the string is as follows:
argv[0] = the name of the executable file. For example, change.exe. (This string does not require user input, and is the same as argc, which can be automatically generated by the operating system.
argv[1] = string 1
argv[2] = string 2
argv[3] = string 3
⋮
-
How to input parameters in programming mode?
The platform used isVisualStudio2019, The file to be used isdown.rgb, The file to be generated isup.yuv,cho.rgb,The steps of parameter input are shown in the following figure:
1. Open the properties window of the upper taskbar debugging interface | ![]() |
---|---|
2. Select debug in configuration properties | ![]() |
3. Modify the command parameters as required | ![]() |
Based on the above knowledge, the preliminary realization of color space conversion can be easily carried out. Header filergb2yuv.h,yuv2rgb.hAnd source filesmain.cpp,rgb2yuv.cpp,yuv2rgb.cppcomposition.
Solution Explorer is shown in the figure below:
Experiment codeas follows:
main.cpp
#include <iostream>#include <cstdio>#include <fstream>#include "rgb2yuv.h"#include "yuv2rgb.h"using namespace std;#define size 196608#define usize 65536#define vsize 131072using namespace std;int main(int argc, char** argv){ifstream infile(argv[1],ios::binary);ofstream outYUV(argv[2], ios::binary);ofstream outRGB(argv[3], ios::binary);if (!infile) { cout << "error to open file1!" << endl; }if (!outYUV) { cout << "error to open file2" << endl; }if (!outRGB) { cout << "error to open file3" << endl; }unsigned char* in = new unsigned char[size];unsigned char* YUV = new unsigned char[size];unsigned char* RGB = new unsigned char[size];infile.read((char*)in, size);rgb2yuv(in,YUV,size, usize, vsize);//First conversionyuv2rgb(YUV, RGB, usize, vsize);//The second conversion/*for (int i = 0; i < size; i++){if (abs(in[i] - RGB[i]) > 5)cout << "i=" << i << " in[" << i << "]=" << int(in[i]) << " RGB[" << i << "]=" << int(RGB[i]) << endl;}*/outYUV.write((char*)YUV, size);outRGB.write((char*)RGB, size);delete in;delete YUV;delete RGB;infile.close();outYUV.close();outRGB.close();return 0;}
rgb2yuv.h
#pragma oncevoid rgb2yuv(unsigned char* rgb, unsigned char* yuv, int size,int usize,int vsize);
rgb2yuv.cpp
void rgb2yuv(unsigned char* rgb, unsigned char* yuv,int size,int usize,int vsize){unsigned char r, g, b, y, u, v;int j = 0;for (int i = 0;i < size;){b = *(rgb + i);g = *(rgb + i + 1);r = *(rgb + i + 2);y = ((66 * r + 129 * g + 25 * b) >> 8) + 16;u = ((-38 * r - 74 * g + 112 * b) >> 8) + 128;v = ((112 * r - 94 * g - 18 * b) >> 8) + 128;*(yuv + j) = y;*(yuv + j + usize) = u;*(yuv + j + vsize) = v;i = i + 3;//Each rgb is 1 groupj++;}}
yuv2rgb.h
#pragma oncevoid yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize);
yuv2rgb.cpp
#pragma once#include "yuv2rgb.h"#include <iostream>using namespace std;void yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize){unsigned char r, g, b, y, u, v;int j = 0;for (int i = 0; i < usize; i++){y = *(yuv + i);u = *(yuv + i + usize);v = *(yuv + i + vsize);r = (298 * y + 411 * v - 57344) >> 8;g = (298 * y - 101 * u - 211 * v + 34739) >> 8;b = (298 * y + 519 * u - 71117) >> 8;*(rgb + j) = b;*(rgb + j + 1) = g;*(rgb + j + 2) = r;j = j + 3;}}
Experimental results
down.rgb | up.yuv | cho.rgb |
---|---|---|
![]() | ![]() | ![]() |
among them,down.rgbwithcho.rgbuseYUVviewerPlusThe way to open is:
The opened image is an inverted image (due tobmpThe image format is stored backwards, so.rgbImage usebmpIt will fall when the mode is opened). The pictures in the above table have been rotated using WeChat for easy identification, butYUVFiles andRGBThere is still mirror flipping between files, but it does not affect viewing and comparison.
up.yuvuseYUVviewerPlusThe way to open is:
It can be seen from the comparison chart of the three images,RGBtoYUV'S experiment was successfully completed, andYUVtoRGBThere is a problem with the experiment, which is transferred outcho.rgbThere are more red noise in the image.
Error correction
Inferred available, in progressYUVtoRGBWhen you getRGBThree figures may exceedunsignedcharThe range that the type can represent, that is, possible<0or>255。
So it’s necessary toyuv2rgb.cppThe file is appropriately revised,>255Values are direct=255,<0Values are direct=0。
After modificationyuv2rgb.cppIs as follows:
#pragma once#include "yuv2rgb.h"#include <iostream>using namespace std;void yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize){int r, g, b, y, u, v;int j = 0;for (int i = 0; i < usize; i++){y = int(*(yuv + i));u = int(*(yuv + i + usize));v = int(*(yuv + i + vsize));r = (298 * y + 411 * v - 57344) >> 8;if (r > 255) { r = 255; }if (r < 0) { r = 0; }g = (298 * y - 101 * u - 211 * v + 34739) >> 8;if (g > 255) { g = 255; }if (g < 0) { g = 0; }b = (298 * y + 519 * u - 71117) >> 8;if (b > 255) { b = 255; }if (b < 0) { b = 0; }*(rgb + j) = unsigned char(b);*(rgb + j + 1) = unsigned char(g);*(rgb + j + 2) = unsigned char(r);j = j + 3;}}
Experimental results
down.rgb | up.yuv | cho.rgb |
---|---|---|
![]() | ![]() | ![]() |
So far, almost doneRGBtoYUVwithYUVtoRGBTwo experiments.
Using a lookup table to optimize the code. Header fileyuvrgb.hAnd source filesmain.cpp,yuvrgb.cppcomposition.
Solution Explorer is shown in the figure below:
main.cpp
#include <iostream>#include <cstdio>#include <fstream>#include "yuvrgb.h"using namespace std;#define size 196608#define usize 65536#define vsize 131072#define height 256#define weight 256//Lookup table initializationint* RGBYUV298 = new int[256];int* RGBYUV411 = new int[256];int* RGBYUV101 = new int[256];int* RGBYUV211 = new int[256];int* RGBYUV519 = new int[256];int* RGBYUV66 = new int[256];int* RGBYUV129 = new int[256];int* RGBYUV25 = new int[256];int* RGBYUV38 = new int[256];int* RGBYUV74 = new int[256];int* RGBYUV112 = new int[256];int* RGBYUV94 = new int[256];int* RGBYUV18 = new int[256];int main(int argc, char** argv){initLookupTable();ifstream infile(argv[1],ios::binary);ofstream outYUV(argv[2], ios::binary);ofstream outRGB(argv[3], ios::binary);if (!infile) { cout << "error to open file1!" << endl; }if (!outYUV) { cout << "error to open file2" << endl; }if (!outRGB) { cout << "error to open file3" << endl; }unsigned char* infi = new unsigned char[size];unsigned char* YUVfi = new unsigned char[size];unsigned char* RGBfi = new unsigned char[size];infile.read((char*)infi, size);rgb2yuv(infi, YUVfi, size, usize, vsize);yuv2rgb(YUVfi, RGBfi, usize, vsize);outYUV.write((char*)YUVfi, size);outRGB.write((char*)RGBfi, size);fileend(infi,YUVfi,RGBfi);infile.close();outYUV.close();outRGB.close();return 0;}
yuvrgb.h
#pragma oncevoid yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize);void rgb2yuv(unsigned char* rgb, unsigned char* yuv, int size, int usize, int vsize);void initLookupTable();void fileend(unsigned char* infi, unsigned char* YUVfi, unsigned char* RGBfi);
yuvrgb.cpp
#pragma once#include "yuvrgb.h"#include <iostream>using namespace std;extern int* RGBYUV298;extern int* RGBYUV411;extern int* RGBYUV101;extern int* RGBYUV211;extern int* RGBYUV519;extern int* RGBYUV66 ;extern int* RGBYUV129;extern int* RGBYUV25 ;extern int* RGBYUV38 ;extern int* RGBYUV74 ;extern int* RGBYUV112;extern int* RGBYUV94 ;extern int* RGBYUV18 ;void initLookupTable(){for (int i = 0; i < 256; i++){RGBYUV298[i] = 298 * i;RGBYUV411[i] = 411 * i;RGBYUV101[i] = 101 * i;RGBYUV211[i] = 211 * i;RGBYUV519[i] = 519 * i;RGBYUV66[i] = 66 * i;RGBYUV129[i] = 129 * i;RGBYUV25[i] = 25 * i;RGBYUV38[i] = 38 * i;RGBYUV74[i] = 74 * i;RGBYUV112[i] = 112 * i;RGBYUV94[i] = 94 * i;RGBYUV18[i] = 18 * i;}}void yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize){int r, g, b, y, u, v;int j = 0;for (int i = 0; i < usize; i++){y = int(*(yuv + i));u = int(*(yuv + i + usize));v = int(*(yuv + i + vsize));/*r = (298 * y + 411 * v - 57344) >> 8;*/r = (RGBYUV298[y]+ RGBYUV411[v]-57344)>>8;if (r > 255) { r = 255; }if (r < 0) { r = 0; }/*g = (298 * y - 101 * u - 211 * v + 34739) >> 8;*/g = (RGBYUV298[y] - RGBYUV101[u] - RGBYUV211[v] + 34739) >> 8;if (g > 255) { g = 255; }if (g < 0) { g = 0; }/*b = (298 * y + 519 * u - 71117) >> 8;*/b = (RGBYUV298[y] + RGBYUV519[u] - 71117) >> 8;if (b > 255) { b = 255; }if (b < 0) { b = 0; }*(rgb + j) = unsigned char(b);*(rgb + j + 1) = unsigned char(g);*(rgb + j + 2) = unsigned char(r);j = j + 3;}}void rgb2yuv(unsigned char* rgb, unsigned char* yuv, int size, int usize, int vsize){int r, g, b, y, u, v;int j = 0;for (int i = 0; i < size;){b = int(*(rgb + i));g = int(*(rgb + i + 1));r = int(*(rgb + i + 2));/*y = ((66 * r + 129 * g + 25 * b) >> 8) + 16;*/y = ((RGBYUV66[r] + RGBYUV129[g] + RGBYUV25[b]) >> 8) + 16;/*u = ((-38 * r - 74 * g + 112 * b) >> 8) + 128;*/u = ((-RGBYUV38[r] - RGBYUV74[g] + RGBYUV112[b]) >> 8) + 128;/*v = ((112 * r - 94 * g - 18 * b) >> 8) + 128;*/v = ((RGBYUV112[r] - RGBYUV94[g] - RGBYUV18[b]) >> 8) + 128;/*if ((y > 255) || (u > 255) || (v > 255) || (y < 0) || (u < 0) || (v < 0)){cout << "y=" << y << "u=" << u << "v=" << v << endl;}*/*(yuv + j) = unsigned char(y);*(yuv + j + usize) = unsigned char(u);*(yuv + j + vsize) = unsigned char(v);i = i + 3;//Each rgb is 1 groupj++;}}void fileend(unsigned char* infi, unsigned char* YUVfi, unsigned char* RGBfi){delete infi;delete YUVfi;delete RGBfi;deleteRGBYUV298;deleteRGBYUV411;deleteRGBYUV101;deleteRGBYUV211;deleteRGBYUV519;deleteRGBYUV66;deleteRGBYUV129;deleteRGBYUV25;deleteRGBYUV38;deleteRGBYUV74;deleteRGBYUV112;deleteRGBYUV94;deleteRGBYUV18;}
Experimental results
down.rgb | up.yuv | cho.rgb |
---|---|---|
![]() | ![]() | ![]() |
So far, 4:4:4 is completedRGBFile with 4:4:4YUVConversion between files.
As shown in the figure, it is a 4:2:0 chroma sampling format. So you can get from 4:4:4YUVThe file is converted to 4:2:0YUVThe idea of the document is to keep the originalYWeight,USignal andVThe signals are all odd points in odd rows. So slightly modify the original code, and add a 4:4:4 according to the above ideaYUVFile gets 4:2:0YUVThe function of the file can be realized fromRGBFile to4:2:0YUVFile conversion.
About from4:2:0YUVFile toRGBFile conversion, due to the correlation between pixels, it is conceivable that the originalYWeight,UVThe components can be copied to get the missingUVWeight, re-converted to 4:4:4YUVfile. According to the above ideas, add a 4:2:0YUVFile gets 4:4:4YUVThe function of the file can be realized from4:2:0YUVFile toRGBFile conversion.
The experiment code is as follows:
main.cpp
#include <iostream>#include <cstdio>#include <fstream>#include "yuvrgb.h"using namespace std;#define size 196608#define csize 98304#define usize 65536#define vsize 131072#define height 256#define weight 256//Lookup table initializationint* RGBYUV298 = new int[256];int* RGBYUV411 = new int[256];int* RGBYUV101 = new int[256];int* RGBYUV211 = new int[256];int* RGBYUV519 = new int[256];int* RGBYUV66 = new int[256];int* RGBYUV129 = new int[256];int* RGBYUV25 = new int[256];int* RGBYUV38 = new int[256];int* RGBYUV74 = new int[256];int* RGBYUV112 = new int[256];int* RGBYUV94 = new int[256];int* RGBYUV18 = new int[256];int main(int argc, char** argv){initLookupTable();ifstream infile(argv[1],ios::binary);ofstream outYUV444(argv[2], ios::binary);ofstream outYUV420(argv[3], ios::binary);ofstream outYUV4442(argv[4], ios::binary);ofstream outRGB(argv[5], ios::binary);if (!infile) { cout << "error to open file1!" << endl; }if (!outYUV444) { cout << "error to open file2" << endl; }if (!outYUV420) { cout << "error to open file3" << endl; }if (!outYUV4442) { cout << "error to open file4" << endl; }if (!outRGB) { cout << "error to open file5" << endl; }unsigned char* infi = new unsigned char[size];unsigned char* YUV444fi = new unsigned char[size];unsigned char* YUV420fi = new unsigned char[csize];unsigned char* YUV4442fi = new unsigned char[size];unsigned char* RGBfi = new unsigned char[size];infile.read((char*)infi, size);rgb2yuv(infi, YUV444fi, size, usize, vsize);yuv444Tyuv420(YUV444fi,YUV420fi,size,weight);yuv420Tyuv444(YUV420fi, YUV4442fi, size, weight);yuv2rgb(YUV4442fi, RGBfi, usize, vsize);outYUV444.write((char*)YUV444fi, size);outYUV420.write((char*)YUV420fi, csize);outYUV4442.write((char*)YUV4442fi, size);outRGB.write((char*)RGBfi, csize);fileend(infi,YUV444fi,YUV420fi,RGBfi);infile.close();outYUV444.close();outYUV420.close();outRGB.close();return 0;}
yuvrgb.h
#pragma oncevoid initLookupTable();void yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize);void rgb2yuv(unsigned char* rgb, unsigned char* yuv, int size, int usize, int vsize);void fileend(unsigned char* infi, unsigned char* YUV444fi, unsigned char* YUV420fi, unsigned char* RGBfi);void yuv444Tyuv420(unsigned char* yuv444,unsigned char* yuv420,int size,int weight);void yuv420Tyuv444(unsigned char* YUV420, unsigned char* YUV444, int size, int weight);
yuvrgb.cpp
#pragma once#include "yuvrgb.h"#include <iostream>using namespace std;extern int* RGBYUV298;extern int* RGBYUV411;extern int* RGBYUV101;extern int* RGBYUV211;extern int* RGBYUV519;extern int* RGBYUV66 ;extern int* RGBYUV129;extern int* RGBYUV25 ;extern int* RGBYUV38 ;extern int* RGBYUV74 ;extern int* RGBYUV112;extern int* RGBYUV94 ;extern int* RGBYUV18 ;void initLookupTable(){for (int i = 0; i < 256; i++){RGBYUV298[i] = 298 * i;RGBYUV411[i] = 411 * i;RGBYUV101[i] = 101 * i;RGBYUV211[i] = 211 * i;RGBYUV519[i] = 519 * i;RGBYUV66[i] = 66 * i;RGBYUV129[i] = 129 * i;RGBYUV25[i] = 25 * i;RGBYUV38[i] = 38 * i;RGBYUV74[i] = 74 * i;RGBYUV112[i] = 112 * i;RGBYUV94[i] = 94 * i;RGBYUV18[i] = 18 * i;}}void yuv2rgb(unsigned char* yuv, unsigned char* rgb,int usize,int vsize){int r, g, b, y, u, v;int j = 0;for (int i = 0; i < usize; i++){y = int(*(yuv + i));u = int(*(yuv + i + usize));v = int(*(yuv + i + vsize));r = (RGBYUV298[y]+ RGBYUV411[v]-57344)>>8;if (r > 255) { r = 255; }if (r < 0) { r = 0; }g = (RGBYUV298[y] - RGBYUV101[u] - RGBYUV211[v] + 34739) >> 8;if (g > 255) { g = 255; }if (g < 0) { g = 0; }b = (RGBYUV298[y] + RGBYUV519[u] - 71117) >> 8;if (b > 255) { b = 255; }if (b < 0) { b = 0; }*(rgb + j) = unsigned char(b);*(rgb + j + 1) = unsigned char(g);*(rgb + j + 2) = unsigned char(r);j = j + 3;}}void rgb2yuv(unsigned char* rgb, unsigned char* yuv, int size, int usize, int vsize){int r, g, b, y, u, v;int j = 0;for (int i = 0; i < size;){b = int(*(rgb + i));g = int(*(rgb + i + 1));r = int(*(rgb + i + 2));y = ((RGBYUV66[r] + RGBYUV129[g] + RGBYUV25[b]) >> 8) + 16;u = ((-RGBYUV38[r] - RGBYUV74[g] + RGBYUV112[b]) >> 8) + 128;v = ((RGBYUV112[r] - RGBYUV94[g] - RGBYUV18[b]) >> 8) + 128;*(yuv + j) = unsigned char(y);*(yuv + j + usize) = unsigned char(u);*(yuv + j + vsize) = unsigned char(v);i = i + 3;//Each rgb is 1 groupj++;}}void yuv444Tyuv420(unsigned char* yuv444, unsigned char* yuv420,int size,int weight){int Ysize = size / 3;int j = 0;for (int i = 0; i < Ysize; i++)//Calculation of Y component{*(yuv420 + j) = *(yuv444 + i);j++;}for (int i = 0; i < Ysize; )//Calculation of U component{if (i % 2 == 1){i++;continue;}if ((i) % (2 * weight) == 0){i = i + weight;continue;}*(yuv420 + j) = *(yuv444 + i + Ysize);j++;i++;}for (int i = 0; i < Ysize; )//Calculation of U component{if (i % 2 == 1){i++;continue;}if ((i) % (2 * weight) == 0){i = i + weight;continue;}*(yuv420 + j) = *(yuv444 + i + Ysize + Ysize);j++;i++;}}void yuv420Tyuv444(unsigned char* YUV420, unsigned char* YUV444, int size, int weight){int Ysize = size / 3;int j = 0;for (int i = 0; i < Ysize; i++)//Calculation of Y component{*(YUV444 + i) = *(YUV420 + j);j++;}for (int i = 0; i < Ysize; )//Calculation of U component{if (i % 2 == 1){j = j - 1;*(YUV444 + i + Ysize) = *(YUV420 + j);i++;j++;continue;}if ((i) % (2 * weight) == 0){j = j - weight/2;*(YUV444 + i + Ysize) = *(YUV420 + j);i++;j++;continue;}*(YUV444 + i + Ysize) = *(YUV420 + j);j++;i++;}for (int i = 0; i < Ysize; )//Calculation of U component{if (i % 2 == 1){j = j - 1;*(YUV444 + i + Ysize + Ysize) = *(YUV420 + j);i++;j++;continue;}if ((i) % (2 * weight) == 0){j = j - weight/2;*(YUV444 + i + Ysize + Ysize) = *(YUV420 + j);i++;j++;continue;}*(YUV444 + i + Ysize + Ysize) = *(YUV420 + j);j++;i++;}}void fileend(unsigned char* infi, unsigned char* YUV444fi,unsigned char* YUV420fi, unsigned char* RGBfi){delete infi;delete YUV444fi;delete YUV420fi;delete RGBfi;deleteRGBYUV298;deleteRGBYUV411;deleteRGBYUV101;deleteRGBYUV211;deleteRGBYUV519;deleteRGBYUV66;deleteRGBYUV129;deleteRGBYUV25;deleteRGBYUV38;deleteRGBYUV74;deleteRGBYUV112;deleteRGBYUV94;deleteRGBYUV18;}
Experimental results
The results of the experiment have 5 pictures, the firstdown.rgb, Converted to 4:4:4down444.yuv, And then converted to 4:2:0down420.yuv, And then converted to 4:4:4down4442.yuv, Then turn intocho.rgb. The results are as follows:
down | 444 | 420 | 4442 | cho |
---|---|---|---|---|
![]() | ![]() | ![]() | ![]() | ![]() |
So far, all the work has been completed.
The source of error may be the following reasons:
- RGBFiles andYUVWhen the formula for file conversion is deduced, it has been quantified and rounded off for many times, so that the conversion formula itself has errors.
- When the 4:4:4 sampling format is converted to the 4:2:0 sampling format, more color difference signals are discarded.
- When the 4:2:0 sampling format is converted to the 4:4:4 sampling format, the pixel value of the same point is used to replace the pixel of the missing point.
- In the process of file conversion, data overflow occurred, and the overflow point was changed upward to 0 or downward to 255, which also caused errors.