听说赫夫曼胜过了他的导师,被认为”青出于蓝而胜于蓝“,这句话也是我比较欣赏的,嘻嘻。
了解”赫夫曼树“之前,几个必须要知道的专业名词可要熟练记住啊。
1: 结点的权
“权”就相当于“重要度”,我们形象的用一个具体的数字来表示,然后通过数字的大小来决定谁重要,谁不重要。
2: 路径
树中从“一个结点"到“另一个结点“之间的分支。
3: 路径长度
一个路径上的分支数量。
4: 树的路径长度
从树的根节点到每个节点的路径长度之和。
5: 节点的带权路径路劲长度
其实也就是该节点到根结点的路径长度*该节点的权。
6: 树的带权路径长度
树中各个叶节点的路径长度*该叶节点的权的和,常用WPL(Weight Path Length)表示。
上面说了那么多,肯定是为下面做铺垫,这里说赫夫曼树,肯定是要说赫夫曼树咋好咋好,赫夫曼树是一种最优二叉树,
因为他的WPL是最短的,何以见得?我们可以上图说话。
现在我们做一个WPL的对比:
图A: WPL= 5*2 + 7*2 +2*2+13*2=54
图B:WPL=5*3+2*3+7*2+13*1=48
我们对比一下,图B的WPL最短的,地球人已不能阻止WPL还能比“图B”的小,所以,“图B"就是一颗赫夫曼树,那么大家肯定
要问,如何构建一颗赫夫曼树,还是上图说话。
第一步: 我们将所有的节点都作为独根结点。
第二步: 我们将最小的C和A组建为一个新的二叉树,权值为左右结点之和。
第三步: 将上一步组建的新节点加入到剩下的节点中,排除上一步组建过的左右子树,我们选中B组建新的二叉树,然后取权值。
第四步: 同上。
大家都知道,字符,汉字,数字在计算机中都是以0,1来表示的,相应的存储都是有一套编码方案来支撑的,比如ASC码。
这样才能在"编码“和”解码“的过程中不会成为乱码,但是ASC码不理想的地方就是等长的,其实我们都想用较少的空间来存储
更多的东西,那么我们就要采用”不等长”的编码方案来存储,那么“何为不等长呢“?其实也就是出现次数比较多的字符我们采用短编码,
出现次数较少的字符我们采用长编码,恰好,“赫夫曼编码“就是不等长的编码。
这里大家只要掌握赫夫曼树的编码规则:左子树为0,右子树为1,对应的编码后的规则是:从根节点到子节点
A: 111
B: 10
C: 110
D: 0
不知道大家懂了没有,不懂的话多看几篇,下面说下赫夫曼的具体实现。
第一步:构建赫夫曼树。
第二步:对赫夫曼树进行编码。
第三步:压缩操作。
第四步:解压操作。
1:首先看下赫夫曼树的结构,这里字段的含义就不解释了。
#region 赫夫曼树结构 /// <summary> /// 赫夫曼树结构 /// </summary> public class HuffmanTree { public int weight { get; set; } public int parent { get; set; } public int left { get; set; } public int right { get; set; } } #endregion
2: 创建赫夫曼树,原理在上面已经解释过了,就是一步一步的向上搭建,这里要注意的二个性质定理:
当叶子节点为N个,则需要N-1步就能搭建赫夫曼树。
当叶子节点为N个,则赫夫曼树的节点总数为:(2*N)-1个。
#region 赫夫曼树的创建 /// <summary> /// 赫夫曼树的创建 /// </summary> /// <param name="huffman">赫夫曼树</param> /// <param name="leafNum">叶子节点</param> /// <param name="weight">节点权重</param> public HuffmanTree[] CreateTree(HuffmanTree[] huffman, int leafNum, int[] weight) { //赫夫曼树的节点总数 int huffmanNode = 2 * leafNum - 1; //初始化节点,赋予叶子节点值 for (int i = 0; i < huffmanNode; i++) { if (i < leafNum) { huffman[i].weight = weight[i]; } } //这里面也要注意,4个节点,其实只要3步就可以构造赫夫曼树 for (int i = leafNum; i < huffmanNode; i++) { int minIndex1; int minIndex2; SelectNode(huffman, i, out minIndex1, out minIndex2); //最后得出minIndex1和minindex2中实体的weight最小 huffman[minIndex1].parent = i; huffman[minIndex2].parent = i; huffman[i].left = minIndex1; huffman[i].right = minIndex2; huffman[i].weight = huffman[minIndex1].weight + huffman[minIndex2].weight; } return huffman; } #endregion #region 选出叶子节点中最小的二个节点 /// <summary> /// 选出叶子节点中最小的二个节点 /// </summary> /// <param name="huffman"></param> /// <param name="searchNodes">要查找的结点数</param> /// <param name="minIndex1"></param> /// <param name="minIndex2"></param> public void SelectNode(HuffmanTree[] huffman, int searchNodes, out int minIndex1, out int minIndex2) { HuffmanTree minNode1 = null; HuffmanTree minNode2 = null; //最小节点在赫夫曼树中的下标 minIndex1 = minIndex2 = 0; //查找范围 for (int i = 0; i < searchNodes; i++) { ///只有独根树才能进入查找范围 if (huffman[i].parent == 0) { //如果为null,则认为当前实体为最小 if (minNode1 == null) { minIndex1 = i; minNode1 = huffman[i]; continue; } //如果为null,则认为当前实体为最小 if (minNode2 == null) { minIndex2 = i; minNode2 = huffman[i]; //交换一个位置,保证minIndex1为最小,为后面判断做准备 if (minNode1.weight > minNode2.weight) { //节点交换 var temp = minNode1; minNode1 = minNode2; minNode2 = temp; //下标交换 var tempIndex = minIndex1; minIndex1 = minIndex2; minIndex2 = tempIndex; continue; } } if (minNode1 != null && minNode2 != null) { if (huffman[i].weight <= minNode1.weight) { //将min1临时转存给min2 minNode2 = minNode1; minNode1 = huffman[i]; //记录在数组中的下标 minIndex2 = minIndex1; minIndex1 = i; } else { if (huffman[i].weight < minNode2.weight) { minNode2 = huffman[i]; minIndex2 = i; } } } } } } #endregion3:对哈夫曼树进行编码操作,形成一套“模板”,效果跟ASC模板一样,不过一个是不等长,一个是等长。
#region 赫夫曼编码 /// <summary> /// 赫夫曼编码 /// </summary> /// <param name="huffman"></param> /// <param name="leafNum"></param> /// <param name="huffmanCode"></param> public string[] HuffmanCoding(HuffmanTree[] huffman, int leafNum) { int current = 0; int parent = 0; string[] huffmanCode = new string[leafNum]; //四个叶子节点的循环 for (int i = 0; i < leafNum; i++) { //单个字符的编码串 string codeTemp = string.Empty; current = i; //第一次获取最左节点 parent = huffman[current].parent; while (parent != 0) { //如果父节点的左子树等于当前节点就标记为0 if (current == huffman[parent].left) codeTemp += "0"; else codeTemp += "1"; current = parent; parent = huffman[parent].parent; } huffmanCode[i] = new string(codeTemp.Reverse().ToArray()); } return huffmanCode; } #endregion
4:模板生成好了,我们就要对指定的测试数据进行压缩处理
#region 对指定字符进行压缩 /// <summary> /// 对指定字符进行压缩 /// </summary> /// <param name="huffmanCode"></param> /// <param name="alphabet"></param> /// <param name="test"></param> public string Encode(string[] huffmanCode, string[] alphabet, string test) { //返回的0,1代码 string encodeStr = string.Empty; //对每个字符进行编码 for (int i = 0; i < test.Length; i++) { //在模版里面查找 for (int j = 0; j < alphabet.Length; j++) { if (test[i].ToString() == alphabet[j]) { encodeStr += huffmanCode[j]; } } } return encodeStr; } #endregion
5: 数据进行还原操作。
#region 对指定的二进制进行解压 /// <summary> /// 对指定的二进制进行解压 /// </summary> /// <param name="huffman"></param> /// <param name="leafNum"></param> /// <param name="alphabet"></param> /// <param name="test"></param> /// <returns></returns> public string Decode(HuffmanTree[] huffman, int huffmanNodes, string[] alphabet, string test) { string decodeStr = string.Empty; //所有要解码的字符 for (int i = 0; i < test.Length; ) { int j = 0; //赫夫曼树结构模板(用于循环的解码单个字符) for (j = huffmanNodes - 1; (huffman[j].left != 0 || huffman[j].right != 0); ) { if (test[i].ToString() == "0") { j = huffman[j].left; } if (test[i].ToString() == "1") { j = huffman[j].right; } i++; } decodeStr += alphabet[j]; } return decodeStr; } #endregion
最后上一下总的运行代码
using System; using System.Collections.Generic; using System.Linq; using System.Text; namespace HuffmanTree { class Program { static void Main(string[] args) { //有四个叶节点 int leafNum = 4; //赫夫曼树中的节点总数 int huffmanNodes = 2 * leafNum - 1; //各节点的权值 int[] weight = { 5, 7, 2, 13 }; string[] alphabet = { "A", "B", "C", "D" }; string testCode = "DBDBDABDCDADBDADBDADACDBDBD"; //赫夫曼树用数组来保存,每个赫夫曼都作为一个实体存在 HuffmanTree[] huffman = new HuffmanTree[huffmanNodes].Select(i => new HuffmanTree() { }).ToArray(); HuffmanTreeManager manager = new HuffmanTreeManager(); manager.CreateTree(huffman, leafNum, weight); string[] huffmanCode = manager.HuffmanCoding(huffman, leafNum); for (int i = 0; i < leafNum; i++) { Console.WriteLine("字符:{0},权重:{1},编码为:{2}", alphabet[i], huffman[i].weight, huffmanCode[i]); } Console.WriteLine("原始的字符串为:" + testCode); string encode = manager.Encode(huffmanCode, alphabet, testCode); Console.WriteLine("被编码的字符串为:" + encode); string decode = manager.Decode(huffman, huffmanNodes, alphabet, encode); Console.WriteLine("解码后的字符串为:" + decode); } } #region 赫夫曼树结构 /// <summary> /// 赫夫曼树结构 /// </summary> public class HuffmanTree { public int weight { get; set; } public int parent { get; set; } public int left { get; set; } public int right { get; set; } } #endregion /// <summary> /// 赫夫曼树的操作类 /// </summary> public class HuffmanTreeManager { #region 赫夫曼树的创建 /// <summary> /// 赫夫曼树的创建 /// </summary> /// <param name="huffman">赫夫曼树</param> /// <param name="leafNum">叶子节点</param> /// <param name="weight">节点权重</param> public HuffmanTree[] CreateTree(HuffmanTree[] huffman, int leafNum, int[] weight) { //赫夫曼树的节点总数 int huffmanNode = 2 * leafNum - 1; //初始化节点,赋予叶子节点值 for (int i = 0; i < huffmanNode; i++) { if (i < leafNum) { huffman[i].weight = weight[i]; } } //这里面也要注意,4个节点,其实只要3步就可以构造赫夫曼树 for (int i = leafNum; i < huffmanNode; i++) { int minIndex1; int minIndex2; SelectNode(huffman, i, out minIndex1, out minIndex2); //最后得出minIndex1和minindex2中实体的weight最小 huffman[minIndex1].parent = i; huffman[minIndex2].parent = i; huffman[i].left = minIndex1; huffman[i].right = minIndex2; huffman[i].weight = huffman[minIndex1].weight + huffman[minIndex2].weight; } return huffman; } #endregion #region 选出叶子节点中最小的二个节点 /// <summary> /// 选出叶子节点中最小的二个节点 /// </summary> /// <param name="huffman"></param> /// <param name="searchNodes">要查找的结点数</param> /// <param name="minIndex1"></param> /// <param name="minIndex2"></param> public void SelectNode(HuffmanTree[] huffman, int searchNodes, out int minIndex1, out int minIndex2) { HuffmanTree minNode1 = null; HuffmanTree minNode2 = null; //最小节点在赫夫曼树中的下标 minIndex1 = minIndex2 = 0; //查找范围 for (int i = 0; i < searchNodes; i++) { ///只有独根树才能进入查找范围 if (huffman[i].parent == 0) { //如果为null,则认为当前实体为最小 if (minNode1 == null) { minIndex1 = i; minNode1 = huffman[i]; continue; } //如果为null,则认为当前实体为最小 if (minNode2 == null) { minIndex2 = i; minNode2 = huffman[i]; //交换一个位置,保证minIndex1为最小,为后面判断做准备 if (minNode1.weight > minNode2.weight) { //节点交换 var temp = minNode1; minNode1 = minNode2; minNode2 = temp; //下标交换 var tempIndex = minIndex1; minIndex1 = minIndex2; minIndex2 = tempIndex; continue; } } if (minNode1 != null && minNode2 != null) { if (huffman[i].weight <= minNode1.weight) { //将min1临时转存给min2 minNode2 = minNode1; minNode1 = huffman[i]; //记录在数组中的下标 minIndex2 = minIndex1; minIndex1 = i; } else { if (huffman[i].weight < minNode2.weight) { minNode2 = huffman[i]; minIndex2 = i; } } } } } } #endregion #region 赫夫曼编码 /// <summary> /// 赫夫曼编码 /// </summary> /// <param name="huffman"></param> /// <param name="leafNum"></param> /// <param name="huffmanCode"></param> public string[] HuffmanCoding(HuffmanTree[] huffman, int leafNum) { int current = 0; int parent = 0; string[] huffmanCode = new string[leafNum]; //四个叶子节点的循环 for (int i = 0; i < leafNum; i++) { //单个字符的编码串 string codeTemp = string.Empty; current = i; //第一次获取最左节点 parent = huffman[current].parent; while (parent != 0) { //如果父节点的左子树等于当前节点就标记为0 if (current == huffman[parent].left) codeTemp += "0"; else codeTemp += "1"; current = parent; parent = huffman[parent].parent; } huffmanCode[i] = new string(codeTemp.Reverse().ToArray()); } return huffmanCode; } #endregion #region 对指定字符进行压缩 /// <summary> /// 对指定字符进行压缩 /// </summary> /// <param name="huffmanCode"></param> /// <param name="alphabet"></param> /// <param name="test"></param> public string Encode(string[] huffmanCode, string[] alphabet, string test) { //返回的0,1代码 string encodeStr = string.Empty; //对每个字符进行编码 for (int i = 0; i < test.Length; i++) { //在模版里面查找 for (int j = 0; j < alphabet.Length; j++) { if (test[i].ToString() == alphabet[j]) { encodeStr += huffmanCode[j]; } } } return encodeStr; } #endregion #region 对指定的二进制进行解压 /// <summary> /// 对指定的二进制进行解压 /// </summary> /// <param name="huffman"></param> /// <param name="leafNum"></param> /// <param name="alphabet"></param> /// <param name="test"></param> /// <returns></returns> public string Decode(HuffmanTree[] huffman, int huffmanNodes, string[] alphabet, string test) { string decodeStr = string.Empty; //所有要解码的字符 for (int i = 0; i < test.Length; ) { int j = 0; //赫夫曼树结构模板(用于循环的解码单个字符) for (j = huffmanNodes - 1; (huffman[j].left != 0 || huffman[j].right != 0); ) { if (test[i].ToString() == "0") { j = huffman[j].left; } if (test[i].ToString() == "1") { j = huffman[j].right; } i++; } decodeStr += alphabet[j]; } return decodeStr; } #endregion } }