实验内容:
1.DPCM编解码原理
DPCM是差分预测编码调制的缩写,是比较典型的预测编码系统。在DPCM系统中,需要注意的是预测器的输入是已经解码以后的样本。之所以不用原始样本来做预测,是因为在解码端无法得到原始样本,只能得到存在误差的样本。因此,在DPCM编码器中实际内嵌了一个解码器,如编码器中虚线框中所示。
在一个DPCM系统中,有两个因素需要设计:预测器和量化器。理想情况下,预测器和量化器应进行联合优化。实际中,采用一种次优的设计方法:分别进行线性预测器和量化器的优化设计。
2.DPCM编码系统的设计
在本次实验中,我们采用固定预测器和均匀量化器。预测器采用左侧、上方预测均可。量化器采用8比特均匀量化。本实验的目标是验证DPCM编码的编码效率。首先读取一个256级的灰度图像,采用自己设定的预测方法计算预测误差,并对预测误差进行8比特均匀量化(基本要求)
。还可对预测误差进行1比特、2比特和4比特的量化设计。
在DPCM编码器实现的过程中可同时输出预测误差图像和重建图像。将预测误差图像写入文件并将该文件输入Huffman编码器,得到输出码流、给出概率分布图并计算压缩比。将原始图像文件输入Huffman编码器,得到输出码流、给出概率分布图并计算压缩比。最后比较两种系统(1.DPCM+熵编码和2.仅进行熵编码)之间的编码效率(压缩比和图像质量)。压缩质量以PSNR进行计算。
代码实现
#include <iostream>
#include<fstream>
using namespace std;
int main()
{
ifstream originalFile("down.yuv", ios::binary);
ofstream rebuildFile("rebuild8.yuv", ios::binary);
ofstream differenceFile("difference8.yuv", ios::binary);
if (!originalFile)
{
cout << "Cannot open originalFile !" << endl;
}
else
{
cout << "originalFile opened successfully !" << endl;
}
if (!rebuildFile)
{
cout << "Cannot open rebuildFile !" << endl;
}
else
{
cout << "rebuildFile opened successfully !" << endl;
}
if (!differenceFile)
{
cout << "Cannot open differenceFile !" << endl;
}
else
{
cout << "differenceFile opened successfully !" << endl;
}
int width = 256;
int height = 256;
int size = width * height;
unsigned char* yBuffer = new unsigned char[size];
unsigned char* uvBuffer = new unsigned char[size * 2];
originalFile.read((char*)yBuffer, size);
originalFile.read((char*)uvBuffer, size * 2);
unsigned char* previous = new unsigned char[size];
unsigned char* rebuildDifference = new unsigned char[size];
unsigned char* quantify = new unsigned char[size];
int bit = 8;
for (int i = 0; i < height; i++)
{
previous[i * width] = yBuffer[i * width];
}
int x = 0;
for (int i = 0; i < height; i++)
{
for (int j = 0; j < width; j++)
{
x = yBuffer[i * width + j] - previous[i * width + j];
if (x < 0)
quantify[i * width + j] = 128 - (int)(abs(x) / pow(2, 9 - bit) + 0.5);
else
quantify[i * width + j] = 128 + (int)(abs(x) / pow(2,9-bit) + 0.5);
previous[i * width + j + 1] = previous[i * width + j] + (quantify[i * width + j] - 128) * 2;
}
}
rebuildFile.write((char*)previous, size);
rebuildFile.write((char*)uvBuffer, size * 2);
for (int i = 0; i < size; i++)
{
rebuildDifference[i] = yBuffer[i] - previous[i] + 128;
}
for (int i = 0; i < size * 2; i++)
{
uvBuffer[i] = 128;
}
differenceFile.write((char*)rebuildDifference, size);
differenceFile.write((char*)uvBuffer, size * 2);
//信号噪声比计算
double PNSR = 0;
double mse = 0;
for (int i = 0; i < height; i++)
{
for (int j = 0; j < width; j++)
mse += pow((yBuffer[i * width + j] - previous[i * width + j]), 2);
}
mse = mse / (width * height);
PNSR = 10 * log10(pow((pow(2, 8) - 1), 2) / mse);
cout << bit << "bit quantify PSNR = " << PNSR << "\n";
delete[]yBuffer;
delete[]uvBuffer;
delete[]previous;
delete[]rebuildDifference;
delete[]quantify;
originalFile.close();
rebuildFile.close();
differenceFile.close();
return 0;
}
输出结果:重建图像以及预测误差图像
8bit
4bit
2bit
1bit
信号噪声比计算:
信噪比越高,图像质量越好;量化比越高的图像质量越好。
预测误差概率分布:
两种系统之间的比较(8bit)