[精品]影像压缩概论.ppt

  1. 1、本文档共36页,可阅读全部内容。
  2. 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
  3. 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载
  4. 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
查看更多
[精品]影像压缩概论

MPEG4 影像壓縮概論 丁文初 2005/12/15 Outline Introduction Image video fundamentals Concept of MPEG4 compression Q A Digital video compression(1) Digital video for storage, broadcasting, data processing, low cost, VLSI implementation…… Uncompressed data: VCD: 3.62 MB/s = 218 MB/min (352x240) DVD: 14.5 MB/s = 870 MB/min (704x480) Compressed: VCD: 0.13 KB/s =7.8 MB/min DVD: 0.9 MB/s = 54 MB/min Digital video compression(2) Transmission bandwidth is limited 2G (GSM): 9.6kbps 2.5G (GPRS): 64kbps 3G (WCDMA, CDMA2000): 2Mbps ADSL: 1Mbps ~ 12Mbps (downstream) 64bps ~ 1Mbps (upstream) WLAN: 54Mbps Image fundamentals (color space) Y (Luma) Cb, Cr (Chroma) Pixel (畫素) Picture resolution (解析度) 800 x 600, 1024x768, 720x480…… VCD: 352x240, DVD:704x480 手機30萬畫素 ? 640x480 = 307,200 For example: R(8-bit), G(8-bit), B(8-bit) ? 1 pixel 352x240x3 = 253440 (bytes) Sampling (取樣) Sampling (取樣) 4:1:1 YCbCr (DV) Sampling (取樣) 4:2:0 YCbCr History MPEG (Moving Picture Expert Group) ITU (International Telecommunication Union) ISO (International Standards Organization) Basic video compressing flow Intra V.S. Inter Basic frame type General MPEG encoder architecture Abbreviation DCT (Discrete Cosine Transform) Q (Quantization) IDCT (Inverse DCT), IQ (Inverse Q) ME (Motion Estimation) MC (Motion Compensation) VLC (Variable Length codes) Processing unit: block 1 block = 8 x 8 (pixel) 1 macro block (MB) = 16 x 16 (pixel) DCT (Discrete Cosine Transform) Compaction: The significant coefficients are clustered around the DC coefficient. De-correlation: The insignificant coefficients are close to zero and can be discard without affecting image quality. DCT basis patterns DCT example (only take 6 of 64 coefficients to reconstruct) DCT coefficient Q (Quantization) Divide all DCT coefficients by integer (quantization table) Lossy compression Only few and significant coefficients are left Quantization example ? After Q DC/AC prediction Zig-zag scan Non-zero coefficients are grouped together. VL

文档评论(0)

138****7331 + 关注
实名认证
内容提供者

该用户很懒,什么也没介绍

1亿VIP精品文档

相关文档