ABCs.425.09.4up - Todays Lecture DSP ABCs: JPEG, JPEG-2K,...

Info iconThis preview shows pages 1–4. Sign up to view the full content.

View Full Document Right Arrow Icon
Hemami JPEG lec.-1-12/2/09 DSP ABCs: JPEG, JPEG-2K, MPEG, MP3, AAC Multimedia Compression Standards ECE 425 Fall 2009 Prof. Hemami Hemami JPEG lec.-2-12/2/09 Today’s Lecture Define compression. Baseline JPEG. JPEG-2000. MPEG MP3 & AAC Hemami JPEG lec.-3-12/2/09 Compression Goal: represent a signal with fewer bits than are required in “raw” form. Examples: MP3 JPEG (images) / MPEG (video) / JBIG (fax) gzip Another way to word this: Eliminate the redundancy in the data to provide an efficient representation that preserves only the essential information. How do we define “redundancy”? Sounds the same? Or IS the same? Hemami JPEG lec.-4-12/2/09 Two Regimes for Compression Lossless compression permits an exact recovery of the original signal. Important for computer files, not so important for audio, images, video, . .. Lossy compression does not allow recovery of the original signal. The amount of loss is controllable, though. ..it can be unnoticeable (MP3s when played in a car), or very noticeable (very low rate JPEG image so you see the blocks).
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Hemami JPEG lec.-5-12/2/09 Generic Lossy Signal Compression 1. Data transform is invertible and most commonly a frequency transform because the visual/audio systems have frequency-selective channels. .. Block-based (e.g., Baseline JPEG w/DCT; audio coding). Full-signal-based (e.g., JPEG-2K w/ wavelets). Data Transform Quantization Entropy Code 1 2 3 Hemami JPEG lec.-6-12/2/09 Lossy Signal Compression, ctd. 2. Quantization reduces the amount of data (this provides the “loss”). 3. Entropy coding is lossless and compacts the bit stream. Some really efficient coding techniques can omit this block and still get very good compression performance. Data Transform Quantization Entropy Code 1 2 3 Hemami JPEG lec.-7-12/2/09 Baseline JPEG The data transform is the Discrete Cosine Transform II. The “loss” is provided by quantization of the DCT coefficients. The quantizer step sizes are hopefully selected based on the human visual system and your desired compression ratio. The entropy coding technique can be either Huffman or arithmetic. 88 × Hemami JPEG lec.-8-12/2/09 Baseline JPEG image component Q 8x8 DCT Quantize DPCM Zig-Zag Scan Run-length encode DC AC Entropy Code 001 0001 (3,14) (3,21) ... arrange into bitstream
Background image of page 2
Hemami JPEG lec.-9-12/2/09 Coding the DC and AC Coefficients Quantized DC coefficients are predictively coded — instead of coding the DC coefficients themselves, code . This reduces the dynamic range of the DC coefficients. AC coefficients are quantized and then run- length-encoded into pairs (number of zeroes preceding non-zero quantized coefficient, quantized coefficient). dn () xn xn 1 = Hemami JPEG lec.-10-12/2/09 Quantization After the DCT, we still have numbers to represent a block: .
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 4
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 01/16/2010 for the course ECE 4250 at Cornell University (Engineering School).

Page1 / 17

ABCs.425.09.4up - Todays Lecture DSP ABCs: JPEG, JPEG-2K,...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online