ADV601
–8–
REV. 0
Table II. ADV601 Typical Quantization of Mallat Data Block
Data
1
Mallat Bin Width Reciprocal Bin
Blocks Factors Width Factors
39 0x007F 0x0810
40 0x009A 0x06a6
41 0x009A 0x06a6
36 0x00BE 0x0564
33 0x00BE 0x0564
30 0x00E4 0x047e
34 0x00E6 0x0474
35 0x00E6 0x0474
37 0x00E6 0x0474
38 0x00E6 0x0474
31 0x0114 0x03b6
32 0x0114 0x03b6
27 0x0281 0x0199
24 0x0281 0x0199
21 0x0301 0x0155
25 0x0306 0x0153
26 0x0306 0x0153
28 0x0306 0x0153
29 0x0306 0x0153
22 0x03A1 0x011a
23 0x03A1 0x011a
5 0x0A16 0x0066
18 0x0A16 0x0066
12 0x0C1A 0x0055
20 0x0C2E 0x0054
19 0x0C2E 0x0054
17 0x0C2E 0x0054
16 0x0C2E 0x0054
14 0x0E9D 0x0046
13 0x0E9D 0x0046
6 0x1DDC 0x0022
9 0x1DDC 0x0022
3 0x23D5 0x001d
11 0x2410 0x001c
10 0x2410 0x001c
8 0x2410 0x001c
7 0x2410 0x001c
5 0x2B46 0x0018
4 0x2B46 0x0018
0 0xA417 0x0006
2 0xC62B 0x0005
1 0xC62B 0x0005
NOTE
1
The Mallat block numbers, Bin Width factors, and Reciprocal Bin Width
factors in Table II correspond to the shading per-cent fill) of Mallat blocks in
Figure 8.
THE RUN LENGTH CODER AND HUFFMAN CODER
This block contains two types of entropy coders that achieve
mathematically loss-less compression: run-length and Huffman.
The run-length coder looks for long strings of zeros and replaces
it with short hand symbols. Table III illustrates an example of
how compression is possible.
The Huffman coder is a digital compressor/decompressor that
can be used for compressing any type of digital data. Essentially,
an ideal Huffman coder creates a table of the most commonly
occurring code sequences (typically zero and small values near
zero) and then replaces those codes with some shorthand. The
ADV601 employs three fixed Huffman tables; it does not create
tables.
The filters and the quantizer increase the number of zeros and
strings of zeros, which improves the performance of the entropy
coders. The higher the selected compression ratio, the more
zeros and small value sequences the quantizer needs to generate.
The transformed image in Figure 5 shows that the filter bank
concentrates zeros and small values in the higher frequency
blocks.
Encoding vs. Decoding
The decoding of compressed video follows the exact path as
encoding but in reverse order. There is no need to calculate Bin
Widths during decode because the Bin Width is stored in the
compressed image during encode.
PROGRAMMER’S MODEL
A host device configures the ADV601 using the Host I/O Port.
The host reads from status registers and writes to control regis-
ters through the Host I/O Port.
An optional DSP can perform Bin Width calculations for the
ADV601. The ADV601 can transfer data from component
video statistics registers and receive data for Bin Width registers
as a packet I/O using the DSP I/O Port. Table IV illustrates the
format used to describe the ADV601’s read and write registers.
Table IV. Register Description Conventions
Register Name
Register Type (Indirect or Direct, Read or Write) and Address
Register Functional Description Text
Bit [#] or Bit or Bit Field Name and Usage Description
Bit Range
[High:Low]
0 Action or Indication When Bit Is Cleared (Equals 0)
1 Action or Indication When Bit Is Set (Equals 1)
Table III. Uncompressed Versus Compressed Data Using Run-Length Coding
0000000000000000000000000000000000000000000000000000000000000000000(uncompressed)
57 Zeros (Compressed)
