
ZR36050
August 1993
ZORAN Corporation
I
1705 Wyatt Drive
I
Santa Clara, CA 95054
I
(408) 986-1314
I
FAX (408) 986-1240
JPEG IMAGE COMPRESSION PROCESSOR
ADVANCE INFORMATION
I
Implements JPEG Baseline image compression and
expansion, including:
- DCT/IDCT operations
- Quantization
- Variable length coding/decoding
Full support of the JPEG Baseline standard, including:
- Bit and byte stuffing
- JPEG markers including restart (RST), application (APP),
and comment (COM)
JPEG Lossless compression and expansion
DMA/SLAVE bus interface
Motion video (30 frames/sec) compression/expansion
capability for CCIR resolution (720 x 480)
I
I
I
I
FEATURES
I
“Fast Preview” option
- Preview of “thumbnail” version of images (up to 25x faster)
Bit rate control option
- Guarantees compressed image file size
Low cost solution
- Low cost single chip
- Support for inexpensive memories
- Requires minimal host intervention
TTL compatible
27 and 21 MSamples/sec data-rate
Standby mode for very low power consumption
100-pin plastic quad flat-pack (PQFP) packaging
I
I
I
I
I
I
I
I
I
I
Computer and multimedia add-in boards
Full-motion video compression/expansion
Digital still cameras and peripherals
Security and industrial systems
APPLICATIONS
GENERAL DESCRIPTION
The ZR36050 is a high-speed JPEG Image Compression Pro-
cessor that performs the algorithm specified by the JPEG
Baseline and JPEG Lossless standards for high-quality image
compression and expansion of continuous-tone color or mono-
chrome images. The ZR36050 performs Discrete Cosine
Transform (DCT), quantization and variable-length encoding for
image compression (coding), and the corresponding inverse
operations for expansion (decoding).
In the JPEG Baseline encoding operation, the ZR36050
performs the DCT operation on 8 x 8 blocks of image data, con-
verting image data into its spatial frequency components, and
quantizes them using a user defined “quantization table.”
Because the human visual system is less sensitive at the higher
spatial frequencies, these higher frequency components can be
quantized more coarsely than the lower-frequency components,
with negligible effect on image quality.
The coarser quantization of high-frequency coefficients results in
long strings of zero valued quantized coefficients, when the 8x8
blocks are scanned in zigzag order. The scanned coefficients
are characterized in terms of their nonzero values and the zero
run lengths. As a result, a long string of zeroes is coded as a
single number. The ZR36050 then performs Huffman coding
using user-defined Huffman tables, whereby bit patterns of dif-
ferent lengths code the nonzero values (values that occur
frequently use the shortest codes; while those that infrequently
occur use the longest codes). These techniques greatly reduce
the amount of memory needed to store an image.
In the decoding operation, the compressed data is decoded (the
inverse of the Huffman and the zigzag modified-run-length
coding), and dequantized. A 2-D inverse Discrete Cosine Trans-
form is performed on the DCT coefficients, resulting in an
expanded image.
Figure 1. ZR36050 Logical Pinout
CWE
CCS
CAEN
CBUSY
WR
CS
INT
DINT
DREQ
DACK
12
PIXEL
DSYNC
EOS
COMP
RESET
STDBY
FREEZE
END
CL
CLKEN
STOP
ZR36050
JPEG IMAGE
COMPRESSION
PROCESSOR
Host Interface
Compressed
Data Interface
PIXEL
Interface
Control
2
CLK_IN
COE
CODE
DATA
ADDR
VSS
+5V
CSYNC
RD
DCT
Coefficient
Output
Clock
COEF
11
8
8
10
I
I
I
I
Videophones and color FAX machines
Color printers and scanners
Fixed bit rate image transmission devices
Cost-sensitive image compression systems
This document was created with FrameMaker 4.0.4