ZR36050 Просмотр технического описания (PDF) - Unspecified

Номер в каталоге

Компоненты Описание

Список матч

ZR36050

JPEG IMAGE COMPRESSION PROCESSOR

Unspecified 

ADVANCE INFORMATION

ZR36050

Baseline Compression Pass, and the resulting compressed data

is fully JPEG compatible. At the completion of the Compression

Pass with or without Bit Rate Control, the ZR36050 calculates a

New Scale Factor (NSF) and saves it in the SF Internal Memory

register. The NSF can be used in the next encoding operation or

the host can overwrite it by its own Scale Factor.

Auto Bit Rate Control

In this mode the ZR36050 performs a Statistical Pass followed

automatically (without host intervention) by a Compression Pass

with Bit Rate Control. The ZR36050 computes the new Scale

Factor and the Allocation Factor at the end of the Statistical

Pass, and rescales the quantization tables by the new scale

factor at the start of the Compression Pass. The DQT marker

must be enabled for Auto Bit Rate Control to work correctly.

Tables-Only Pass

In this mode, the ZR36050 generates compressed data in the

JPEG abbreviated format for table specification. The abbreviat-

ed format compressed data contains only the SOI marker,

quantization and/or Huffman tables specifications (DQT and/or

DHT marker segments), optional APP, COM marker segments,

and the EOI marker. The content of the MARKERS_EN register

specifies which marker segments are to be included in the

abbreviated format data. The Pixel Interface is inoperative in this

mode.

Tables Preload for Encoding

Prior to encoding a sequence of images with the same quantiza-

tion and/or Huffman tables, this mode is used to preload the

tables. The DQTI and DHTI bits of the MARKERS_EN register

specify which tables to preload. If DQTI is set, the ZR36050 mul-

tiplies the quantization tables, specified in the DQT segment of

Internal Memory by the Scale Factor, and stores the scaled

tables in the Quantization Tables Store. If DHTI is set, it decodes

the Huffman tables specifications from the DHT segment of

Internal Memory, where they are specified in accordance with

the JPEG syntax, and stores the decoded tables in the Huffman

Tables Store, for use in compressing the images. In this mode,

the Pixel Interface and the Code Buffer are inoperative.

JPEG Lossless Compression

In a Compression Pass, if the ZR36050 finds the SOF3 frame

marker in the Internal Memory, it switches to the JPEG Lossless

Compression mode. JPEG Lossless compression uses a spatial

algorithm, so the DCT Unit is bypassed in this mode. No quanti-

zation is performed, so the Quantization Tables Store is not

used. The ZR36050 encodes the image samples using the

JPEG one-dimensional horizontal prediction method and

Huffman coding. Up to two Huffman tables are allowed in JPEG

Lossless compression. The sample precision can be from 2 to 12

bits. Horizontally subsampled components are supported and no

point transform is performed (the point transform parameter in

the scan header marker segment is ignored).

DECODING MODES

JPEG Baseline Expansion

In Expansion mode, the ZR36050 reads compressed data and

expands it using the inverse of the JPEG Baseline encoding

algorithm. It stores the marker segments extracted from the com-

pressed data in the Marker Segments Section of Internal

Memory, where the host can access the information. When it

encounters a DQT or DHT marker segment, it decodes the

segment and stores the tables in the Table Stores. It expands

any subsequent compressed image data using the stored tables.

If the compressed data contains multiple instances of a marker

segment, a new segment will overwrite the previous segment of

the same type in the internal memory. This mode is also used to

decode abbreviated format tables-only compressed data.

Fast Preview

In this mode, the ZR36050 decodes only the DC coefficients

from the JPEG Baseline compressed image data, and outputs

them, after level-shifting to form unsigned DC values, via the

Pixel Interface. Since only one sample is output for each 8 x 8

block, the result is a thumbnail version of the image, scaled down

by a factor of eight horizontally and vertically. It is generated up

to 25 times faster than full image expansion. The inverse DCT

computation is bypassed in this mode.

Tables Preload for Decoding

The ZR36050 reads the quantization and/or Huffman tables from

Internal Memory and stores them in the Table Stores for

decoding images in the JPEG abbreviated data-only format. The

DQTI and DHTI bits of the MARKERS_EN register specify which

tables to preload. If DQTI is set, the ZR36050 multiplies the

quantization tables by the Scale Factor and stores the scaled

tables in the Quantization Tables Store. If DHTI is set, it decodes

the Huffman tables specifications from the DHT segment of the

Internal Memory, where they are specified in accordance with

the JPEG syntax, and stores the decoded tables in the Huffman

Tables Store. In this mode, the Pixel Interface and the Code

Buffer are inoperative.

JPEG Lossless Expansion

In Expansion mode, if the ZR36050 detects the SOF3 frame

marker in the compressed data, it switches to the JPEG Lossless

Expansion mode. The DCT Unit is bypassed in this mode, and

the quantization tables are not used. The compressed image is

decoded by the inverse of the method used in the Lossless Com-

pression mode. As in Lossless Compression, up to two Huffman

tables are allowed. The sample precision can be from 2 to 12

bits. Horizontally subsampled components are supported and no

point transform is performed (the point transform parameter in

the scan header marker segment is ignored. Therefore if a

lossless bitstream with point transform other than zero is input

into the ZR36050, it will be decoded as if the point transform is

zero).

12

Share Link: