首先介绍以下什么是哈夫曼树(来自百度百科)
哈夫曼树─即最优二叉树,带权路径长度最小的二叉树,经常应用于数据压缩。 在计算机信息处理中,“哈夫曼编码”是一种一致性编码法(又称“熵编码法”),用于数据的无损耗压缩。这一术语是指使用一张特殊的编码表将源字符(例如某文件中的一个符号)进行编码。这张编码表的特殊之处在于,它是根据每一个源字符出现的估算概率而建立起来的(出现概率高的字符使用较短的编码,反之出现概率低的则使用较长的编码,这便使编码之后的字符串的平均期望长度降低,从而达到无损压缩数据的目的)。
构造哈夫曼树的主要思想:
构造哈夫曼树非常简单,将所有的节点放到一个队列中,用一个节点替换两个频率最低的节点,新节点的频率就是这两个节点的频率之和。这样,新节点就是两个被替换节点的父节点了。如此循环,直到队列中只剩一个节点(树根)。
这里用到了最小优先队列。
我这里用STL来实现。(这里有优先队列的介绍)
template<typename T>
struct cmp
{
bool operator()(TreeNode<T>* t1, TreeNode<T>* t2)
{
return !(*t1 < *t2);
}
};
优先队列的定义:
priority_queue<TreeNode*,vector<TreeNode* >,cmp > pri_que;
哈夫曼树节点结构
template<typename T>
class TreeNode
{
public:
TreeNode():pfather(NULL),plchild(NULL),prchild(NULL)
{
}
T data;
TreeNode *pfather;
TreeNode *plchild;
TreeNode *prchild;
bool operator < (const TreeNode& rhs)
{
return data < rhs.data;
}
};
构造哈夫曼树
每次从最小优先队列取头两个节点,合并后放回最小优先队列,如此重复。
template<typename T>
TreeNode<T>* MakeHuffmanTree(T* begin, T* end) //构造哈夫曼树
{
priority_queue<TreeNode<T>*,vector<TreeNode<T>* >,cmp<T> > pri_que;
T *iter = begin;
TreeNode<T>* pNode;
TreeNode<T>* pf = NULL;
while(iter != end)
{
pNode = new TreeNode<T>;
pNode->data = *iter++;
pNode->pfather = pf;
pri_que.push(pNode);
}
TreeNode<T>* plchild;
TreeNode<T>* prchild;
while(pri_que.size() > 1)//取两个小的合并
{
plchild = pri_que.top();
pri_que.pop();
prchild = pri_que.top();
pri_que.pop();
pNode = new TreeNode<T>;
pNode->plchild = plchild;
pNode->prchild = prchild;
pNode->data = plchild->data + prchild->data;
pri_que.push(pNode);
}
pNode = pri_que.top();
pri_que.pop();
return pNode;
}
构造哈夫曼树这个函数的参数是一个结构体,保存着对应字符,其频率,编码值。
重载它的+运算符,为了合并时的+运算(只是频率相加)。
到此为止,已经可以把哈夫曼树生成出来了。
template<typename T>
struct mydata
{
mydata(){}
mydata(int i):freq(i)
{
}
string coded;
int freq;
T data;
bool operator<(const mydata& rhs)
{
return freq < rhs.freq;
}
mydata operator+(mydata& rhs)
{
return mydata(freq + rhs.freq);
}
};
我们可以通过DFS将每个叶子节点的路径记录下来(用一个全局变量数组path),然后得到它的编码。
当发现当前节点是叶子节点,就把当前的路径赋值至该叶子节点的编码属性(coded)。
const int MAXLEN = 20;
char path[MAXLEN] = {0};
template<typename T>
void DFS(T* root,int deep = -1, char a = '-') //DFS 得到叶子节点的编码
{
if(root == NULL)
return;
if(a != '-')
path[deep] = a;
if(root->plchild == NULL || root->prchild == NULL)//leaf
(root->data).coded = string(path,path + deep + 1);
if(root->plchild != NULL)
DFS(root->plchild, deep + 1, '0');
if(root->prchild != NULL)
DFS(root->prchild, deep + 1, '1');
}
这样整个哈夫曼编码工作已经完成,为了查看我们的编码结果,我们可以用BFS跟DFS来看到我们的结果。在这里我选择了BFS。
当遍历到叶子节点,就将其编码属性(coded)和其对应字符输出。
template<typename T,typename U>
void BFS(T* root, mydata<U>* data) //BFS 将叶子节点的编码,提到data指向的数据
{
queue<T*> que;
que.push(root);
T* pT = NULL;
while(!que.empty())
{
pT = que.front();
//cout<<pT->data.freq<<endl;
que.pop();
if(pT->plchild != NULL)
que.push(pT->plchild);
if(pT->prchild != NULL)
que.push(pT->prchild);
if(pT->plchild == NULL || pT->prchild == NULL)// leaf 提取叶子节点的编码
{
//cout<<(pT->data).data<<":"<<(pT->data).coded<<endl;
mydata<U>* pd = data;
while((pT->data).data != pd->data)
pd++;
assert(pd->data == (pT->data).data);
pd->coded = (pT->data).coded;
}
}
}
测试驱动代码
mydata<char> *pdata = new mydata<char>[4];
pdata[0].data = 'a';
pdata[0].freq = 7;
pdata[1].data = 'b';
pdata[1].freq = 5;
pdata[2].data = 'c';
pdata[2].freq = 2;
pdata[3].data = 'd';
pdata[3].freq = 4;
TreeNode<mydata<char> >* pihuffmanTree = MakeHuffmanTree(pdata, pdata + 4);
DFS(pihuffmanTree);
BFS(pihuffmanTree);
为了更方便的使用我将这些封装到一个类里面。
template<typename T>
class Huffman
{
public:
void Coded(string& coded);//传入待输出的编码
void DeCode(const string& codedstr,string& decodestr);//输入待解码字符串,输出解码字符串
void InputData(T* begin,T* end);//传入数据
private:
string FindVal(char c);
void m_CalcFreq(T* begin, T* end);//计算输入数据的频率
TreeNode<mydata<T> > *root;//huffman根节点
mydata<T>* data;
int data_size;
T* m_begin;//保存原始数据的开始与结束的位置
T* m_end;
//string codedstr;
};
输入数据并计算其频率。
用map容器来统计输入字符每个出现的个数。
template<typename T>
void Huffman<T>::InputData(T* begin, T* end)
{
this->m_begin = begin;
this->m_end = end;
m_CalcFreq(begin, end);
}
template<typename T>
void Huffman<T>::m_CalcFreq(T* begin, T* end)
{
int len = end - begin;
//data_size = len;
if(len == 0)
return;
map<T,int> countMap;
map<T,int>::iterator mapIter = countMap.begin();
T *pT = begin;
while(pT != end)
{
mapIter = countMap.find(*pT);
if(mapIter != countMap.end())//在map里有没有字符*iter
++mapIter->second;
else
{
countMap.insert(make_pair(*pT,1));
}
pT++;
}
data_size = countMap.size();
data = new mydata<T>[data_size];
int i = 0;
for (mapIter = countMap.begin(); mapIter != countMap.end(); ++mapIter)
{
data[i].data = mapIter->first;
data[i].freq = mapIter->second;
i++;
}
}
编码
template<typename T>
void Huffman<T>::Coded(string& coded)
{
root = MakeHuffmanTree(data,data + data_size);
DFS(root);
BFS(root,data);
cout<<"code:"<<endl;
for (int i = 0; i < data_size; ++i)
{
cout<<data[i].data<<":"<<data[i].coded<<endl;
}
T *begin = m_begin;
while (begin != m_end)
{
coded += FindVal(*begin);
begin++;
}
//string subcode =
}
解码
template<typename T>
void Huffman<T>::DeCode(const string& codedstr,string& decodestr)
{
string::const_iterator iter = codedstr.begin();
TreeNode<mydata<T> >* curNode = root;
while (iter != codedstr.end())
{
if (curNode->plchild == NULL || curNode->prchild == NULL)
{
decodestr += (curNode->data).data;
curNode = root;
continue;
}
if (*iter == '0')
curNode = curNode->plchild;
if(*iter == '1')
curNode = curNode->prchild;
iter++;
}
}
测试驱动程序
char *pmystr = "cbcddddbbbbaaaaaaa";
Huffman<char> h;
h.InputData(pmystr, pmystr + 18);
cout<<"originstr: "<<pmystr<<endl;
string coded;
h.Coded(coded);
cout<<"coded: "<<coded<<endl;
string decode;
h.DeCode(coded,decode);
cout<<"decode: "<<decode<<endl;
完整程序(环境:VS2012)
#include <iostream>
//#include <algorithm>
#include <queue>
#include <string>
#include <vector>
#include <cassert>
#include <map>
using namespace std;
template<typename T>
class TreeNode
{
public:
TreeNode():pfather(NULL),plchild(NULL),prchild(NULL)
{
}
T data;
TreeNode *pfather;
TreeNode *plchild;
TreeNode *prchild;
bool operator < (const TreeNode& rhs)
{
return data < rhs.data;
}
};
template<typename T>
struct cmp
{
bool operator()(TreeNode<T>* t1, TreeNode<T>* t2)
{
return !(*t1 < *t2);
}
};
template<typename T>
TreeNode<T>* MakeHuffmanTree(T* begin, T* end) //构造哈夫曼树
{
priority_queue<TreeNode<T>*,vector<TreeNode<T>* >,cmp<T> > pri_que;
T *iter = begin;
TreeNode<T>* pNode;
TreeNode<T>* pf = NULL;
while(iter != end)
{
pNode = new TreeNode<T>;
pNode->data = *iter++;
pNode->pfather = pf;
pri_que.push(pNode);
}
TreeNode<T>* plchild;
TreeNode<T>* prchild;
while(pri_que.size() > 1)//取两个小的合并
{
//cout<<static_cast<TreeNode<T>* >(pri_que.top())->data<<endl;
//pri_que.pop();
plchild = pri_que.top();
pri_que.pop();
prchild = pri_que.top();
pri_que.pop();
pNode = new TreeNode<T>;
pNode->plchild = plchild;
pNode->prchild = prchild;
pNode->data = plchild->data + prchild->data;
pri_que.push(pNode);
}
pNode = pri_que.top();
pri_que.pop();
return pNode;
}
template<typename T>
struct mydata
{
mydata(){}
mydata(int i):freq(i)
{
}
string coded;
int freq;
T data;
bool operator<(const mydata& rhs)
{
return freq < rhs.freq;
}
mydata operator+(mydata& rhs)
{
return mydata(freq + rhs.freq);
}
};
template<typename T,typename U>
void BFS(T* root, mydata<U>* data) //BFS 将叶子节点的编码,提到data指向的数据
{
queue<T*> que;
que.push(root);
T* pT = NULL;
while(!que.empty())
{
pT = que.front();
//cout<<pT->data.freq<<endl;
que.pop();
if(pT->plchild != NULL)
que.push(pT->plchild);
if(pT->prchild != NULL)
que.push(pT->prchild);
if(pT->plchild == NULL || pT->prchild == NULL)// leaf 提取叶子节点的编码
{
//cout<<(pT->data).data<<":"<<(pT->data).coded<<endl;
mydata<U>* pd = data;
while((pT->data).data != pd->data)
pd++;
assert(pd->data == (pT->data).data);
pd->coded = (pT->data).coded;
}
}
}
const int MAXLEN = 20;
char path[MAXLEN] = {0};
template<typename T>
void DFS(T* root,int deep = -1, char a = '-') //DFS 得到叶子节点的编码
{
if(root == NULL)
return;
if(a != '-')
path[deep] = a;
if(root->plchild == NULL || root->prchild == NULL)//leaf
(root->data).coded = string(path,path + deep + 1);
if(root->plchild != NULL)
DFS(root->plchild, deep + 1, '0');
if(root->prchild != NULL)
DFS(root->prchild, deep + 1, '1');
}
template<typename T>
class Huffman
{
public:
void Coded(string& coded);
void DeCode(const string& codedstr,string& decodestr);
void InputData(T* begin,T* end);
private:
string FindVal(char c);
void m_CalcFreq(T* begin, T* end);
TreeNode<mydata<T> > *root;
mydata<T>* data;
int data_size;
T* m_begin;
T* m_end;
//string codedstr;
};
template<typename T>
void Huffman<T>::InputData(T* begin, T* end)
{
this->m_begin = begin;
this->m_end = end;
m_CalcFreq(begin, end);
}
template<typename T>
void Huffman<T>::m_CalcFreq(T* begin, T* end)
{
int len = end - begin;
//data_size = len;
if(len == 0)
return;
map<T,int> countMap;
map<T,int>::iterator mapIter = countMap.begin();
T *pT = begin;
while(pT != end)
{
mapIter = countMap.find(*pT);
if(mapIter != countMap.end())
++mapIter->second;
else
{
countMap.insert(make_pair(*pT,1));
}
pT++;
}
data_size = countMap.size();
data = new mydata<T>[data_size];
int i = 0;
for (mapIter = countMap.begin(); mapIter != countMap.end(); ++mapIter)
{
data[i].data = mapIter->first;
data[i].freq = mapIter->second;
i++;
}
}
template<typename T>
void Huffman<T>::Coded(string& coded)
{
root = MakeHuffmanTree(data,data + data_size);
DFS(root);
BFS(root,data);
cout<<"code:"<<endl;
for (int i = 0; i < data_size; ++i)
{
cout<<data[i].data<<":"<<data[i].coded<<endl;
}
T *begin = m_begin;
while (begin != m_end)
{
coded += FindVal(*begin);
begin++;
}
//string subcode =
}
template<typename T>
void Huffman<T>::DeCode(const string& codedstr,string& decodestr)
{
string::const_iterator iter = codedstr.begin();
TreeNode<mydata<T> >* curNode = root;
while (iter != codedstr.end())
{
if (curNode->plchild == NULL || curNode->prchild == NULL)
{
decodestr += (curNode->data).data;
curNode = root;
continue;
}
if (*iter == '0')
curNode = curNode->plchild;
if(*iter == '1')
curNode = curNode->prchild;
iter++;
}
}
template<typename T>
string Huffman<T>::FindVal(char c)
{
for (int i = 0; i < data_size ; ++i)
{
if (c != data[i].data)
continue;
return data[i].coded;
}
return string();
}
int main()
{
//mydata<char> *pdata = new mydata<char>[4];
//pdata[0].data = 'a';
//pdata[0].freq = 7;
//pdata[1].data = 'b';
//pdata[1].freq = 5;
//pdata[2].data = 'c';
//pdata[2].freq = 2;
//pdata[3].data = 'd';
//pdata[3].freq = 4;
////int a[12]={14,10,56,7,83,22,36,91,3,47,72,0};
//TreeNode<mydata<char> >* pihuffmanTree = MakeHuffmanTree(pdata, pdata + 4);
//DFS(pihuffmanTree);
//BFS(pihuffmanTree);
//string str = "cbcddddbbbbaaaaaaa";
char *pmystr = "cbcddddbbbbaaaaaaa";
Huffman<char> h;
h.InputData(pmystr, pmystr + 18);
cout<<"originstr: "<<pmystr<<endl;
string coded;
h.Coded(coded);
cout<<"coded: "<<coded<<endl;
string decode;
h.DeCode(coded,decode);
cout<<"decode: "<<decode<<endl;
return 0;
}