[IR] Dictionary Coding
Lempel–Ziv–Welch
年发表的论文《A Universal Algorithm for Sequential Data Compression》中提出。
基于统计的数据压缩编码,比如Huffman编码,需要得到先验知识——信源的字符频率,然后进行压缩。但是在大多数情况下,这种先验知识是很难预先获得。
因此,设计一种更为通用的数据压缩编码显得尤为重要。LZ77数据压缩算法应运而生,其核心思想:利用数据的重复结构信息来进行数据压缩。
LZ77: referring to previously processed data as dictionary 利用内部信息作为字典。
在提出基于滑动窗口的LZ77算法后,两位大神Jacob Ziv与Abraham Lempel于年在发表的论文中提出了LZ78算法;
与LZ77算法不同的是LZ78算法使用动态树状词典维护历史字符串。
LZ78: use an explicit dictionary 字典是外置的。
LZ系列压缩算法均为LZ77与LZ78的变种,在此基础上做了优化。
- LZ77:LZSS、LZR、LZB、LZH;
- LZ78:LZW、LZC、LZT、LZMW、LZJ、LZFG。
LZW Encoding:
Video: https://www.youtube.com/watch?v=nW7OARbr7OI
"TO BE OR NOT TO BE OR TO BE OR NOT"
Idea:
以下是我们已知的字典。

再动态补充新发现的pattern字典,从27开始编号,如下所示:
| current | next | code | dictionary | ||
| T | O | 20 | TO | 27 | TO BE OR NOT TO BE OR TO BE OR NOT |
| O | B | 15 | OB | 28 | TO BE OR NOT TO BE OR TO BE OR NOT |
| B | E | 2 | BE | 29 | TO BE OR NOT TO BE OR TO BE OR NOT |
| E | O | 5 | EO | 30 | TO BE OR NOT TO BE OR TO BE OR NOT |
| O | R | 15 | OR | 31 | TO BE OR NOT TO BE OR TO BE OR NOT |
| R | N | 18 | RV | 32 | TO BE OR NOT TO BE OR TO BE OR NOT |
| N | O | 14 | NO | 33 | TO BE OR NOT TO BE OR TO BE OR NOT |
| O | T | 15 | OT | 34 | TO BE OR NOT TO BE OR TO BE OR NOT |
| T | T | 20 | TT | 35 | TO BE OR NOT TO BE OR TO BE OR NOT |
| TO | B | 27 | TOB | 36 | TO BE OR NOT TO BE OR TO BE OR NOT |
| BE | O | 29 | BEO | 37 | TO BE OR NOT TO BE OR TO BE OR NOT |
| OR | T | 31 | ORT | 38 | TO BE OR NOT TO BE OR TO BE OR NOT |
| TOB | E | 36 | TOBE | 39 | TO BE OR NOT TO BE OR TO BE OR NOT |
| EO | R | 30 | EOR | 40 | TO BE OR NOT TO BE OR TO BE OR NOT |
| RN | O | 32 | RNO | 41 | TO BE OR NOT TO BE OR TO BE OR NOT |
| OT | # | 34 | N/A | N/A | TO BE OR NOT TO BE OR TO BE OR NOT |
| Input | Output | ||||
这里共16行,也就是原来的24字节 --> 16字节。
LZW Decoding:
| code | prev | output | dictionary | ||
| 20 | T | TO BE OR NOT TO BE OR TO BE OR NOT | |||
| 15 | T | O | TO | 27 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 2 | O | B | OB | 28 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 5 | B | E | BE | 29 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 15 | E | O | EO | 30 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 18 | O | R | OR | 31 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 14 | R | N | RN | 32 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 15 | N | O | NO | 33 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 20 | O | T | OT | 34 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 27 | T | TO | TT | 35 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 29 | TO | BE | TOB | 36 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 31 | BE | OR | BEO | 37 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 36 | OR | TOB | ORT | 38 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 30 | TOB | EO | TOBE | 39 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 32 | EO | RN | EOR | 40 | TO BE OR NOT TO BE OR TO BE OR NOT |
| 34 | RN | OT | RNO | 41 | TO BE OR NOT TO BE OR TO BE OR NOT |
| <Output> | <Input> | ||||
可见与encoding时表格一一对应的关系。
就是还原表格的过程。
[IR] Dictionary Coding的更多相关文章
- [IR] Huffman Coding
为了保证:Block中,所有的叶子在所有的中间结点的前面.Static: Huffman coding Dynamic: Adaptive Huffman 一些概念 压缩指标 • Compress a ...
- [IR] Arithmetic Coding
Statistical methods的除了huffman外的另一种常见压缩方式. Huffman coding的非连续数值特性成为了无法达到香农极限的先天无法弥补的缺陷,但Arithmetic co ...
- 本人AI知识体系导航 - AI menu
Relevant Readable Links Name Interesting topic Comment Edwin Chen 非参贝叶斯 徐亦达老板 Dirichlet Process 学习 ...
- [Code] 烧脑之算法模型
把博客的算法过一遍,我的天呐多得很,爱咋咋地! 未来可考虑下博弈算法. 基本的编程陷阱:[c++] 面试题之犄角旮旯 第壹章[有必要添加Python] 基本的算法思想:[Algorithm] 面试题之 ...
- HANA Architecture
1 HANA 是基于内存计算的.行列都支持.使用列存储,列存储的特点是高压缩,查询快,节约空间, ---SAP HANA supports both, but is particularly opti ...
- python之最强王者(8)——字典(dictionary)
1.Python 字典(Dictionary) 字典是另一种可变容器模型,且可存储任意类型对象. 字典的每个键值(key=>value)对用冒号(:)分割,每个对之间用逗号(,)分割,整个字典包 ...
- Dictionary Learning(字典学习、稀疏表示以及其他)
第一部分 字典学习以及稀疏表示的概要 字典学习(Dictionary Learning)和稀疏表示(Sparse Representation)在学术界的正式称谓应该是稀疏字典学习(Sparse Di ...
- Coding源码学习第一部分(AppDelegate.m)
前言:在此首先感谢开源,感谢大神们的无私分享. Coding 的主页:https://coding.net/app#app-feature Coding 自己家的仓库:https://coding.n ...
- [IR] Compression
关系:Vocabulary vs. collection size Heaps’ law: M = kTbM is the size of the vocabulary, T is the numbe ...
随机推荐
- iOS for MachineLearning
链接: 手把手教你在应用里用上iOS机器学习框架Core ML iOS11 新功能开发之 - "高大上"的 CoreML 与 Vision Core ML介绍 (Apple机器学习 ...
- 导出putty配置
原文链接:http://downloadsquad.switched.com/2007/02/01/howto-transfer-your-putty-settings-between-compute ...
- ArcGIS中国工具2.5正式发布
ArcGIS中国工具2.5正式发布 1. 支持1:500,1:1000,1:2000的梯形接幅表和图框打印(见3.2) 2. 支持接幅表,一定间隔如1秒加节点生成并能自动识别打印(见3. ...
- android http中请求访问添加 cookie
第一种HashMap<String, String> map = new HashMap<String, String>();map.put("cookie" ...
- POJ 1595 素数打表水题
[题意简述]:给出N和C,让我们求出N以内的包含N的素数,然后依据若N以内的素数为奇数个,就将中间2*c-1个素数输出:若为偶数个.就将中间2*c个素数输出. [分析]:仅仅要题意理解就简单了. 详见 ...
- .Net Core AES加密解密
一.AES说明 高级加密标准(英语:Advanced Encryption Standard,缩写:AES),在密码学中又称Rijndael加密法,是美国联邦政府采用的一种区块加密标准.这个标准用来替 ...
- pkg-config命令的Makefile.am
举例:通过Makefile调用pkg-config命令. pkg-config - Return metainformation about installed libraries (为了使用lib库 ...
- iSpy免费的开源视频监控平台
iSpy包括英文,Deutsch,Español,Française,Italiano和中文的翻译 iSpy是我们免费的开源视频监控平台.iSpy作为安装的Windows应用程序运行,具有完整的本地用 ...
- open jdk
http://rednaxelafx.iteye.com/blog/1549577 https://www.jianshu.com/p/f98c3acd8df8 http://download.jav ...
- sublime text 3-right click context menu
dd a system wide windows explorer button " Edit with Sublime" similar to how Notepad++ doe ...