《图像处理》课件.pptx
《《图像处理》课件.pptx》由会员分享,可在线阅读,更多相关《《图像处理》课件.pptx(53页珍藏版)》请在淘文阁 - 分享文档赚钱的网站上搜索。
1、数据编码信源编码提高数据表达效率(压缩)信道编码提高数据存贮/传输可靠性一些信号的原始数据量信源信源数据量数据量电话(200Hz 3.4kHz)8k/s 12bits=96kb/s宽带语音(50Hz 7kHz)16k/s 14bits=224kb/s宽带音频(20Hz 20kHz)44.1k/s 2通道16bits=1.412Mb/s图像51251224bits=6Mb视频64048024bits30/s=221Mb/s高清晰度电视128072024bits60/s=1.3Gb/s数据压缩的依据 存在数据冗余时间冗余 时间轴上的相关性空间冗余 像素间的相关性信息熵冗余 数据表示的效率知识冗余
2、先验知识的利用心理视觉冗余 合理的分辨率coding redundancy Redundancy:R=1 L/mApproaches:-Standard Huffman -Shifted Huffman -Arithmetic Coding No redundancy:8bitInterpixel redundancy Spatial redundancyApproaches:-Run Length coding -intraframe predicative coding -LZW codingInterpixel redundancy Spatial,time redundancyAppr
3、oaches:-interframe predicative coding (motion prediction)Psychovisual redundancy Original 25%quality 15%quality 45,339 b 3,905 b 3,197 bRemove Less important informationApproaches:-quantization(lossy),DCT Compression system modelTwo parts:Encoder,DecoderSource encoder:remove redundancyChannel encode
4、r:reduce noiseSource encoder and decoder Mapper:transform input data to a special format for reducing interpixel redundancy数据编码分类无损压缩(lossless):变长编码(Huffman)行程编码算术编码-医学或商用数据的压缩有损数据压缩(lossy)预测编码变换编码(Cosine,Wavelet)-视频数据的压缩Run-Length Encoding(1D)Original data:05,05,05,05,11,11,1A,1A,1A,1A,20,22,22,22,
5、22Encoding:05,04,11,02,1A,04,20,01,22,04Application:be suitable for uniform gray images,FAX codingRun-Length Encoding(2D)05,02,05,01,10,01,21,04,05,01,10,01,10,04,21,02,31,02,10,02,10,02,15,01,31,02,31,01,15,01,31,01Rules:basic unit-maximal square left-right,top-bottom LZW coding根据数据本身包含有重复的字符串(不一定要
6、求空间上相邻)每接收一个字符系列,判断以前是否出现过,然后用出现过的字符串代替该重复的部分,输出对应的“指针”(位置)例如:“aabbccddee”的指针是 0100,以后每次出现该字符串,就用0100代替。LZW NotesDictionary is not null at first,single symbolBuilding decompression dictionary is needed Application:GIF(graphic interchange format)TIFF(tagged image file format)PDF(portable document for
7、mat)Huffman coding灰度 0 1 2 3 4 5 6 7概率 0.05 0.05 0.10 0.15 0.25 0.20 0.15 0.050.100.150.250.300.450.551.001001010101011编码0110001101 10010100110100111编码:从根节点开始到叶子节点!Standard Huffman codingAverage bits:2.2Entropy:2.14 efficiency:0.973Standard Huffman disadvantagesEven if 1 bit error,a series of errors
8、,no solutions (example in next page)When decoding,fail to abstract contents from randomly position Computation:k-2 source rankingTo symbol with less probability:code length is too long(max:k-1 bits)-shifted HuffmanError decoding example of Huffman Code:011,1,01010,0100,01011,00Gray:a1 a2 a3 a4 a5 a6
9、a3,a1,a2,a2,a6.01010 011 1 1 00*Encoding:Gray CodeDecoding:Code Gray010100111100.01010 00 1 1 1 1 00.01010 01011 00.a3,a6,a2,a2,a2,a2,a6.a3,a5,a6.*Add 0:error decoding*011-010,Shifted Huffman encoding 1Truncate:90%(main)-standard HuffmanDivide blocks:assign a prefix to each blockShift:prefix+standar
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 图像处理 图像 处理 课件
限制150内