Upload
piper
View
150
Download
1
Embed Size (px)
DESCRIPTION
静态图像的国际压缩标准 JPEG. 2005 年 fall. 1 概述. JPEG (Joint Photographic Experts Group) 是一个由 ISO 和 IEC 两个组织机构联合组成的一个专家组 1992 年成为国际标准 JPEG 是国际上彩色、灰度、静止图像的第一个国际标准,也是一个适用范围广泛的通用标准。它不仅适用于静止图像的压缩,电视图像序列的帧内图像的压缩编码,也常采用 JPEG 压缩方法。 可用于 多媒体 CD-ROM, 彩色图像传真,图文档案管理等。 该标准制定了有损压缩和无损压缩的编码方案。. - PowerPoint PPT Presentation
Citation preview
JPEG2005fall
1 JPEG(Joint Photographic Experts Group) ISOIEC1992 JPEGJPEGCD-ROM
JPEGCPU
Sequential (line by line) and progressive (successive refinement of the whole image) decoding should both be possible. A lossless, hierarchical coding of the same image with different resolutions should also be possible.The user can select parameters to trade off the quality of the reproduced image, the compression processing time, the size of the compressed image.
The encoded data stream has a fixed interchange formatencoded image datathe chosen parameterstables of the coding process.
DCT
JPEG ModesThe lossy, sequential DCT-based mode (baseline process, base mode)The expanded lossy, DCT-based modeThe lossless modeThe hierarchical mode
10:1--100:1 (40:1)25 JPEG 30:1 1000K33K 1MB
2 An image consists of at least one and at most N=255 components or planes.
Each component is a rectangular array XiYi of pixels (the samples).
The resolution of the individual components may be different.
Each pixel is represented by p bits with values in the range from 0 to 2 p1.All pixels of all components of an image must have the same number of bits.The lossy modes of JPEG use a precision of either 8 or 12 bits per pixel.The lossless modes can use between 2 and 12 bits per pixel. With any other number of bits, the application must suitably transform the image.
The compressed data includes:X (maximum of all Xi)Y (maximum of all Yi)factors Hi and Vi for each plane.Hi : relative horizontal resolutionVi : relative vertical resolution P100
the image is divided into data unitsThe lossless mode uses one pixel as one data unit. The lossy mode uses blocks of 88 pixels.Non-interleaved data ordering
Interleaved data ordering
Interleaved data units of different components are combined into Minimum Coded Units (MCU).
According to the JPEG standarda maximum of four components can be encoded using the interleaved mode. Each MCU can contain at most ten data units.Within an image, some components can be encoded in the interleaved mode and others in the noninterleaved mode.
3 DCT
After image preparation, the uncompressed image samples are grouped into data units of 88 pixelsIn this baseline mode, each sample is encoded using p=8bit. Each pixel is an integer in the range 0 to 255.
Example of DCT and IDCT
DCTDCTPT
The pixel values are shifted into the zero-centered interval (128, 127). 128This transformation must be carried out 64 times per data unit. The result is 64 coefficients. DCT8888
Ti=0,j=0DCACij
DCT88DCT
640480640480,8060JPEG88
88 T
PACAC0PT0ACAC0PAC0ACAC
DCTDCT
88DCT(DCT)
JPEG64DCT() (,)T(,)(,)
T(,)(,)(,)
IDCT
JPEGJPEGYCrCb YCrCb YYCrCb YCrCb .JPEG
JPEGYCrCb YCrCb YYYCrCb YYCrCb JPEG10-310-4JPEG
1-255DCT
DC JPEGDCDPCM()88DC
AC T63(AC)6363(Zig-Zag)Z
ZRLEZ
Z 63AC
DCACentropy codingJPEGDC
For the lossy sequential DCT-based base mode, only Huffman encoding is allowed.In the base mode, two different Huffman tables can be used, one for AC-coefficients and for DC-coefficients.
Example of sequential picture preparation
JPEG
38384 bytes (High quality),11331 bytes (Medium quality), 6968 bytes (Low quality),3687 bytes
JPEGQuality FactorPhotoshop012PhotoPaint2250
maximum medium low 8 3 0 3.7:1 7.1:1 8.8:1
maximum medium low 8 3 0 7.6:1 11.8:1 13.3:1
Factor: 100 Size=326321 bytes Factor: 50 Size=46295 bytes Factor: 25 Size=29360 bytes Factor: 5 Size=9438 bytes
10.6.5 ---JPEG2000 JPEGJPEG88 JPEG JPEG 2000 JPEG JPEG (Discrete Cosine Transform) (Wavelet transform)
JPEG2000 JPEG2000JPEG30%50%JPEG2000JPEG30%50%JPEG2000 JPEGJPEG2000JPEG2000
Progressive Transmission JPEG (Region of Interest) JPEG2000JPEG10%30% JPEG2000JPEG
1910Harr80Stromberg 1984MorletFourier1987MallatMallat
(Wavelet Transformation)Fourier(Mallat)Fourier
48013001InernetCDROM
DSPAnalogDevices/IntelIndeoIndeo
: 22:1
FourierMPEG-4
FourierCTFourier