C++实例解析哈夫曼树
给定n个权值作为n个叶子结点,构造一棵二叉树,若带权路径长度达到最小,称这样的二叉树为最优二叉树,也称为哈夫曼树(Huffman tree)。
1、路径和路径长度
在一棵树中,从一个结点往下可以达到的孩子或子孙结点之间的通路,称为路径。通路中分支的数目称为路径长度。若规定根结点的层数为1,则从根结点到第L层结点的路径长度为L-1。
2、结点的权及带权路径长度
若将树中结点赋给一个有着某种含义的数值,则这个数值称为该结点的权。结点的带权路径长度为:从根结点到该结点之间的路径长度与该结点的权的乘积。
3、树的带权路径长度
树的带权路径长度规定为所有叶子结点的带权路径长度之和,记为WPL
哈夫曼树的构造
哈夫曼树的构造
假设有n个权值,则构造出的哈夫曼树有n个叶子结点。 n个权值分别设为 w1、w2、…、wn,则哈夫曼树的构造规则为:
(1) 将w1、w2、…,wn看成是有n 棵树的森林(每棵树仅有一个结点);
(2) 在森林中选出两个根结点的权值最小的树合并,作为一棵新树的左、右子树,且新树的根结点权值为其左、右子树根结点权值之和;
(3)从森林中删除选取的两棵树,并将新树加入森林;
(4)重复(2)、(3)步,直到森林中只剩一棵树为止,该树即为所求得的哈夫曼树
using namespace std; const int MaxValue = 10000; //初始设定的权值最大值 const int MaxBit = 4; //初始设定的最大编码位数 const int MaxN = 10; //初始设定的最大结点个数 struct HaffNode //哈夫曼树的结点结构 { int weight; //权值 int flag; //标记 int parent; //双亲结点下标 int leftChild; //左孩子下标 int rightChild; //右孩子下标 }; struct Code //存放哈夫曼编码的数据元素结构 { int bit[MaxN]; //数组 int start; //编码的起始下标 int weight; //字符的权值 }; void Haffman(int weight[], int n, HaffNode haffTree[]) //建立叶结点个数为n权值为weight的哈夫曼树haffTree { int j, m1, m2, x1, x2; //哈夫曼树haffTree初始化。n个叶结点的哈夫曼树共有2n-1个结点 for(int i = 0; i < 2 * n - 1 ; i++) { if(i < n) haffTree[i].weight = weight[i]; else haffTree[i].weight = 0; haffTree[i].parent = 0; haffTree[i].flag = 0; haffTree[i].leftChild = -1; haffTree[i].rightChild = -1; } //构造哈夫曼树haffTree的n-1个非叶结点 for(i = 0;i < n-1;i++) { m1 = m2 = MaxValue; x1 = x2 = 0; for(j = 0; j < n+i;j++) { if (haffTree[j].weight < m1 && haffTree[j].flag == 0){ m2 = m1; x2 = x1; m1 = haffTree[j].weight; x1 = j; } else if(haffTree[j].weight < m2 && haffTree[j].flag == 0){ m2 = haffTree[j].weight; x2 = j; } } //将找出的两棵权值最小的子树合并为一棵子树 haffTree[x1].parent = n+i; haffTree[x2].parent = n+i; haffTree[x1].flag = 1; haffTree[x2].flag = 1; haffTree[n+i].weight = haffTree[x1].weight+haffTree[x2].weight; haffTree[n+i].leftChild = x1; haffTree[n+i].rightChild = x2; } } void HaffmanCode(HaffNode haffTree[], int n, Code haffCode[]) //由n个结点的哈夫曼树haffTree构造哈夫曼编码haffCode { Code *cd = new Code; int child, parent; //求n个叶结点的哈夫曼编码 for(int i = 0; i < n; i++) { cd->start = n-1; //不等长编码的最后一位为n-1 cd->weight = haffTree[i].weight; //取得编码对应权值的字符 child = i; parent = haffTree[child].parent; //由叶结点向上直到根结点 while(parent != 0) { if(haffTree[parent].leftChild == child) cd->bit[cd->start] = 0; //左孩子结点编码0 else cd->bit[cd->start] = 1;//右孩子结点编码1 cd->start--; child = parent; parent = haffTree[child].parent; } //保存叶结点的编码和不等长编码的起始位 for(int j = cd->start+1; j < n; j++) haffCode[i].bit[j] = cd->bit[j]; haffCode[i].start = cd->start; haffCode[i].weight = cd->weight; //保存编码对应的权值 } } void main(void){ int i, j, n = 4; int weight[] = {1,3,5,7}; HaffNode *myHaffTree = new HaffNode[2*n+1]; Code *myHaffCode = new Code[n]; if(n > MaxN) { cout << "定义的n越界,修改MaxN! " << endl; exit(0); } Haffman(weight, n, myHaffTree); HaffmanCode(myHaffTree, n, myHaffCode); //输出每个叶结点的哈夫曼编码 for(i = 0; i < n; i++) { cout << "Weight = " << myHaffCode[i].weight << " Code = "; for(j = myHaffCode[i].start+1; j < n; j++) cout << myHaffCode[i].bit[j]; cout << endl; } }