首页 > 其他分享 >[图像处理]YUV图像处理入门3

[图像处理]YUV图像处理入门3

时间:2022-12-17 20:01:48浏览次数:31  
标签:入门 int image unsigned YUV char width 图像处理 input


5 yuv420格式的灰阶测试图

本程序中的函数主要是为YUV420P视频数据流的第一帧图像添加边框。函数的代码如下所示:

/**
* @file 5 yuv_graybar.cpp
* @author luohen
* @brief gray scale bar of yuv
* @date 2018-12-07
*
*/

#include "stdafx.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <iostream>

using namespace std;

/**
* @brief
*
* @param width width of input yuv420p file
* @param height height of input yuv420p file
* @param ymin minimum value of y
* @param ymax maximum value of y
* @param barnum Number of bars
* @param url location of input yuv420p file
* @return int
*/
int yuv420_graybar(int width, int height, int ymin, int ymax, int barnum, const char *url)
{
//每个灰度条的宽度
int barwidth;
//每个灰度阶次范围
float lum_inc;
//计算Y值
unsigned char lum_temp;
//uv分量宽高
int uv_width, uv_height;
//reading yuv image
FILE *input_fp;
if ((input_fp = fopen(url, "rb")) == NULL)
{
printf("%s open error!\n", url);
return -1;
}
else
{
printf("%s open.\n", url);
}
//writing yuv image
FILE *output_fp = fopen("video_result/gray_test.yuv", "wb+");

int t = 0, i = 0, j = 0;

//每个灰度条的宽度
barwidth = width / barnum;
//每个灰度阶次范围
lum_inc = ((float)(ymax - ymin)) / ((float)(barnum - 1));
//uv分量宽高
uv_width = width / 2;
uv_height = height / 2;

unsigned char *data_y = new unsigned char[width * height];
unsigned char *data_u = new unsigned char[uv_width * uv_height];
unsigned char *data_v = new unsigned char[uv_width * uv_height];

//Output Info
//输出信息
printf("Y, U, V value from picture's left to right:\n");
for (t = 0; t < (width / barwidth); t++)
{
//计算Y值
lum_temp = ymin + (char)(t * lum_inc);
printf("%3d, 128, 128\n", lum_temp);
}
//保存数据
for (j = 0; j < height; j++)
{
for (i = 0; i < width; i++)
{
t = i / barwidth;
lum_temp = ymin + (char)(t * lum_inc);
data_y[j * width + i] = lum_temp;
}
}
for (j = 0; j < uv_height; j++)
{
for (i = 0; i < uv_width; i++)
{
data_u[j * uv_width + i] = 128;
}
}
for (j = 0; j < uv_height; j++)
{
for (i = 0; i < uv_width; i++)
{
data_v[j * uv_width + i] = 128;
}
}

fwrite(data_y, width * height, sizeof(unsigned char), output_fp);
fwrite(data_u, uv_width * uv_height, sizeof(unsigned char), output_fp);
fwrite(data_v, uv_width * uv_height, sizeof(unsigned char), output_fp);
fclose(input_fp);
fclose(output_fp);

delete[] data_y;
delete[] data_u;
delete[] data_v;
return 0;
}

/**
* @brief main
*
* @return int
*/
int main()
{
int state = yuv420_graybar(640, 360, 0, 255, 10, "video/graybar.yuv");
return 0;
}

调用函数为:

int yuv420_graybar(int width, int height, int ymin, int ymax, int barnum, const char *url);

实际上这部分代码和前面代码差不多,先取得YUV数据流,类似一个一维数组,读第一帧图像,然后依次读到y,u,v三个分量起始位置,再对y,u,v的像素值分别进行处理。

结果如图所示:

[图像处理]YUV图像处理入门3_i++


6 两张yuv420p图像的峰值信噪比(psnr)计算

本程序中的函数主要是比较两张yuv420p图像的峰值信噪。函数的代码如下所示:

/**
* @file 6 yuv420_psnr.cpp
* @author luohen
* @brief Compute the PSNR values of two yuv files
* @date 2018-12-08
*
*/

#include "stdafx.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <iostream>

using namespace std;

/**
* @brief
*
* @param url1 location of input yuv420p file1
* @param url2 location of input yuv420p file2
* @param w width of input yuv420p file
* @param h height of input yuv420p file
* @return int
*/
int yuv420_psnr(const char *url1, const char *url2, int w, int h)
{
//reading yuv iamges
FILE *fp1 = fopen(url1, "rb+");
FILE *fp2 = fopen(url2, "rb+");

unsigned char *pic1 = new unsigned char[w * h];
unsigned char *pic2 = new unsigned char[w * h];

fread(pic1, 1, w * h, fp1);
fread(pic2, 1, w * h, fp2);

double mse_sum = 0, mse = 0, psnr = 0;
//computing mse
for (int j = 0; j < w * h; j++)
{
mse_sum += pow((double)(pic1[j] - pic2[j]), 2);
}
mse = mse_sum / (w * h);
//computing psnr
psnr = 10 * log10(255.0 * 255.0 / mse);
printf("%5.3f\n", psnr);

delete[] pic1;
delete[] pic2;
fclose(fp1);
fclose(fp2);
return 0;
}

/**
* @brief main
*
* @return int
*/
int main()
{
int state = yuv420_psnr("video/akiyo.yuv", "video/distort_akiyo.yuv", 352, 288);
return 0;
}

调用函数为:

int yuv420_psnr(const char *url1, const char *url2, int w, int h);

这段代码主要是计算两张图像的接近程度,psnr值具体介绍可以见文章:


本文所用的两张图像一张是akiyo视频流首帧图像,另外一张是前面为akiyo加上边框的图像。两张图像的psnr值为13.497。一般psnr值越大两张图像越接近。


7 yuv420图像顺时针旋转90度

本程序中的函数主要是将YUV420P视频数据流的第一帧图像顺时针旋转90度。函数的代码如下所示:

/**
* @file 7 yuv_Rotation90.cpp
* @author luohen
* @brief 90 degree rotation of yuv420 images
* @date 2018-12-08
*
*/

#include "stdafx.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <iostream>

using namespace std;

/**
* @brief Pre-defined image size
*
*/
#define image_h 288
#define image_w 352

/**
* @brief
*
* @param url location of input yuv420p file
* @return int
*/
int yuv420_Rotation90(const char *url)
{
//reading yuv files
FILE *input_fp;
//writingyuv files
FILE *output_fp = fopen("video_result/output_rotation.yuv", "wb+");

//reading yuv datas
if ((input_fp = fopen(url, "rb")) == NULL)
{
printf("%s open error!\n", url);
return -1;
}
else
{
printf("%s open.\n", url);
}

//Input image array definition
unsigned char input_Y[image_h][image_w];
unsigned char input_U[image_h / 2][image_w / 2];
unsigned char input_V[image_h / 2][image_w / 2];

//Output image array definition
unsigned char output_Y[image_w][image_h];
unsigned char output_U[image_w / 2][image_h / 2];
unsigned char output_V[image_w / 2][image_h / 2];

int w = image_w;
int h = image_h;

fread(input_Y, sizeof(unsigned char), w * h, input_fp);
fread(input_U, sizeof(unsigned char), w / 2 * h / 2, input_fp);
fread(input_V, sizeof(unsigned char), w / 2 * h / 2, input_fp);

//Y 90 degree rotation
for (int x = 0; x < h; x++)
{
for (int y = 0; y < w; y++)
{
//旋转之后,输出的x值等于输入的y坐标值
//y值等于输入列高-输入x坐标值-1
output_Y[y][h - x - 1] = input_Y[x][y];
}
}

//u 90 degree rotation
for (int x = 0; x < h / 2; x++)
{
for (int y = 0; y < w / 2; y++)
{
//旋转之后,输出的x值等于输入的y坐标值
//y值等于输入列高-输入x坐标值-1
output_U[y][h / 2 - x - 1] = input_U[x][y];
}
}

//v 90 degree rotation
for (int x = 0; x < h / 2; x++)
{
for (int y = 0; y < w / 2; y++)
{
//旋转之后,输出的x值等于输入的y坐标值
//y值等于输入列高-输入x坐标值-1
output_V[y][h / 2 - x - 1] = input_V[x][y];
}
}

fwrite(output_Y, sizeof(unsigned char), w * h, output_fp);
fwrite(output_U, sizeof(unsigned char), w / 2 * h / 2, output_fp);
fwrite(output_V, sizeof(unsigned char), w / 2 * h / 2, output_fp);

fclose(input_fp);
fclose(output_fp);

return 0;
}

/**
* @brief main
*
* @return int
*/
int main()
{
int state = yuv420_Rotation90("video/akiyo.yuv");
return 0;
}

调用函数为:

int yuv420_Rotation90(const char *url);

这段代码主要是分别提取yuv分量,然后将y,u,v分量分别旋转90度。但是提取yuv分量和以前的代码有所不同。

首先是建立yuv三个分量输入的静态二维数组,相比使用动态数组,这种方式处理数据简单很多,但是需要实现确定输入图像的大小。

unsigned char input_Y[image_h][image_w];

unsigned char input_U[image_h / 2][image_w / 2];

unsigned char input_V[image_h / 2][image_w / 2];

然后建立旋转后的输出数组,输出数组定义是,由于是旋转90度,长宽进行了对调。

unsigned char output_Y[image_w][image_h];

unsigned char output_U[image_w / 2][image_h / 2];

unsigned char output_V[image_w / 2][image_h / 2];

其他旋转操作,就是图像赋值过程。旋转后akiyo图像尺寸变为(288,352)

结果如图所示:

[图像处理]YUV图像处理入门3_ide_02


8 yuv420图像大小重置

本程序中的函数主要是对YUV420P视频数据流的第一帧图像进行缩放或者放大。类似opencv中的resize函数,函数的代码如下所示:

/**
* @file 8 yuv_resize.cpp
* @author luohen
* @brief adjusting yuv image size
* @date 2018-12-08
*
*/

#include "stdafx.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <string.h>
#include <iostream>

using namespace std;

#define HEIGHT 288
#define WIDTH 352

/**
* @brief
*
* @param url location of input yuv420p file
* @param out_width output image width
* @param out_height output image height
* @return int
*/
int yuv420_resize(const char *url, int out_width, int out_height)
{
//input array
unsigned char yin[HEIGHT][WIDTH];
unsigned char uin[HEIGHT / 2][WIDTH / 2];
unsigned char vin[HEIGHT / 2][WIDTH / 2];
//output array
unsigned char *yout = new unsigned char[out_width * out_height];
unsigned char *uout = new unsigned char[out_width / 2 * out_height / 2];
unsigned char *vout = new unsigned char[out_width / 2 * out_height / 2];
///reading yuv file
FILE *input_fp;
//writing yuv file
FILE *output_fp = fopen("video_result/output_resize.yuv", "wb+");

if ((input_fp = fopen(url, "rb")) == NULL)
{
printf("%s open error!\n", url);
return -1;
}
else
{
printf("%s open.\n", url);
}

fread(yin, sizeof(unsigned char), HEIGHT * WIDTH, input_fp);
fread(uin, sizeof(unsigned char), HEIGHT * WIDTH / 4, input_fp);
fread(vin, sizeof(unsigned char), HEIGHT * WIDTH / 4, input_fp);

//Y
for (int i = 0; i < out_height; i++)
{
for (int j = 0; j < out_width; j++)
{
int i_in = round(i * HEIGHT / out_height);
int j_in = round(j * WIDTH / out_width);
yout[i * out_width + j] = yin[i_in][j_in];
}
}

//U
for (int i = 0; i < out_height / 2; i++)
{
for (int j = 0; j < out_width / 2; j++)
{
int i_in = round(i * (HEIGHT / 2) / (out_height / 2));
int j_in = round(j * (WIDTH / 2) / (out_width / 2));
uout[i * out_width / 2 + j] = uin[i_in][j_in];
}
}

//V
for (int i = 0; i < out_height / 2; i++)
{
for (int j = 0; j < out_width / 2; j++)
{
int i_in = round(i * (HEIGHT / 2) / (out_height / 2));
int j_in = round(j * (WIDTH / 2) / (out_width / 2));
vout[i * out_width / 2 + j] = vin[i_in][j_in];
}
}

fwrite(yout, sizeof(unsigned char), out_width * out_height, output_fp);
fwrite(uout, sizeof(unsigned char), out_width * out_height / 4, output_fp);
fwrite(vout, sizeof(unsigned char), out_width * out_height / 4, output_fp);

delete[] yout;
delete[] uout;
delete[] vout;
fclose(input_fp);
fclose(output_fp);

return 0;
}

/**
* @brief main
*
* @return int
*/
int main()
{
int state = yuv420_resize("video/akiyo.yuv", 288, 352);
return 0;
}

调用函数为:

int yuv420_resize(const char *url, int out_width, int out_height);

这段代码也是通过事先设定yuv输入输出的静态二维数组来进行处理的。其中out_width, out_height

是输出图像的宽高,这段代码中输出图像的宽高可以设定为任意值。所用图像resize方法是最简单的最邻近插值法。

插值方法见文章:


当设置调整后的图像宽高为288,352时,结果如下:

[图像处理]YUV图像处理入门3_i++_03

 

标签:入门,int,image,unsigned,YUV,char,width,图像处理,input
From: https://blog.51cto.com/luohenyueji/5950123

相关文章

  • [图像处理]YUV图像处理入门4
    9yuv420图像截取本程序中的函数主要是对YUV420P视频数据流的第一帧图像进行截取。类似opencv中的rect函数,函数的代码如下所示:/***@file9yuv_clip.cpp*@authorluohen......
  • [图像处理]YUV图像处理入门2
    1分离YUV420中YUV分量本程序中的函数主要是将YUV420P视频数据流的第一帧图像中的Y、U、V三个分量分离开并保存成三个文件。函数的代码如下所示:/***@file1yuv_split.cp......
  • [图像处理]YUV图像处理入门1
    目前数字图像处理技术已经应用生活各个方面,但是大部分教程都是利用第三方库(如opencv)对RGB图像格式进行处理。对于YUV图像格式的图像处理教程较少。于是博主搬运总结了多个......
  • Vue项目入门
    1安装VueMac版本安装:https://zhuanlan.zhihu.com/p/435312919Window版本安装:https://blog.csdn.net/weixin_43896253/article/details/116143031开发软件安装:VisualSt......
  • 《小白WEB安全入门》02. 开发篇
    目录初识HTML潜在漏洞初识CSS潜在漏洞初识JS潜在漏洞初识后端潜在漏洞后端能做什么后端种类后端框架潜在漏洞本系列文章只叙述一些超级基础理论知识,极少有实践部分本文......
  • 内存和ssd速度差距(diy从入门到放弃固态能取代内存吗)
    原文地址:https://mpc.7buzou.com/article/5745.html虽然内存和SSD都是存储产品,但两者的待遇却完全不一样:8G内存可以轻松卖到200元的价格,而200元可以买到256GB的SSD,价格差......
  • 微信小程序入门记录
    0.基础认识本质上就是仅支持微信环境的网页,其使用的wxml即html,wxss即css和js即js。但也由于微信api的加持使得开发更为简单(注册登录和支付等api)。注意:小程序的逻辑层和......
  • 【图像处理笔记】小波变换
     【图像处理笔记】总目录0引言曾经有人问我有关haar的东西,我没答上来,耿耿于怀,所以我从傅里叶变换学到小波变换再到haar小波,蓦然回首,才发现他当时问的是haar特征。但是,......
  • STL入门
    目的:复用性提升,为了建立数据结构和算法的一套标准。STL简介:STL:标准模板库STL广义:容器,算法,迭代器容器和算法之间通过迭代器进行无缝衔接STL几乎所有代码都采用了模板类......
  • VSCode编辑器极简使用入门
    VSCode(VisualStudioCode)是一款开源、跨平台、轻量级的代码编辑器,具有非常丰富的插件生态。他本身就是JavaScript+Electron(/ɪˈlektrɒn/电子)代码开发的。官方下载......