ADV601LCJST Analog Devices Inc, ADV601LCJST Datasheet - Page 30

• Manufacturer Part Number: ADV601LCJST
• Description: IC CODEC VIDEO DSP/SRL 120LQFP
• Manufacturer: Analog Devices Inc
• Type: Video Codecr

Specifications of ADV601LCJST

• Rohs Status: RoHS non-compliant
• Data Interface: DSP, Serial
• Resolution (bits): 8 b
• Sigma Delta: No
• Voltage - Supply, Analog: 4.5 V ~ 5.5 V
• Voltage - Supply, Digital: 4.5 V ~ 5.5 V
• Operating Temperature: 0°C ~ 70°C
• Mounting Type: Surface Mount
• Package / Case: 120-LQFP

ADV601LC
GETTING THE MOST OUT OF ADV601LC
The unique sub-band block structure of luminance and color
components in the ADV601LC offers many unique application
benefits. Analog Devices will offer a Feature Software Library as
well as separate feature application documentation to help users
exploit these features. The following section provides an over-
view of only some of the features and how they are achieved with
the ADV601LC. Please refer to Figures 2 and 3 as necessary.
Higher Compression With Interfield Techniques
The ADV601LC normally operates as a field-independent
codec. However, through use of the sub-bands it is possible to
use the ADV601LC with interfield techniques to achieve even
higher levels of compression. In such applications, each field is
not compressed separately, thus accessing the compressed bit
stream can only be done at specific points in time. There are
two general ways this can be accomplished:
• Subsampling high frequency blocks
• Updating the image with motion detection
Scalable Compression Technology
The ADV601LC offers many different options for scaling the
image, the compressed bit stream bandwidth and the processing
horsepower for encode or decode. Because the ADV601LC
employs decimators, interpolators and filters in the filter bank,
the scaling function creates much higher quality images than
achieved through pixel dropping. Mixing and matching the
many scaling options is useful in network applications where
transmission pipes may vary in available bit rate, and decode/
encode capabilities may be a mix of software and hardware.
These are the key options:
• Extract scaled images by factors of 2 from the compressed bit stream
• Use software to decode bit stream
The human visual system is more sensitive to interframe
motion of low frequency block than to motion in high fre-
quency blocks. The host software driver of the ADV601LC
allows exploitation of this option to achieve higher com-
pression. Note that the compressed bit stream can only be
accessed at points where the high frequency blocks have just
been updated.
In applications where the video is likely to have no motion for
extended periods of time (video surveillance in a vacant build-
ing, for instance), it is only necessary to update the image
either periodically or when motion occurs. By using the wave-
let sub-bands to detect motion (see later in this section), it is
possible to achieve very high levels of compression when
motion is infrequent.
This is useful in video editing applications where thumbnail
sketches of fields need to be displayed. In this case, editing
software can quickly extract and decode the desired image.
This technique eliminates the burden of decoding an entire
image and then scaling to the desired size.
Decoding an entire CCIR-601 resolution image in real time at
50/60 fields per second does require the ADV601LC hard-
ware. Analog Devices provides a bit-exact ADV601LC
simulator that can decode a scaled image in real time or a full-
size image off-line. Image size and frame rates depend on the
performance of the host processor.
–30–
• Scale bit stream
• Use software to encode
Parametric Image Filtering
The ADV601LC offers a unique set of image filtering capa-
bilities not found in other compression technologies. The
ADV601LC quantizer is capable of attenuating any or all of the
luminance or chrominance blocks during encode or decode.
Here are some of the possible applications:
• Parametric softening of color saturation and contrast during encode
• Color saturation control
• Contrast control
• Fade to black
Mixing of Two or More Images
Blocks from different images can be mixed into the bit stream
and then sent to the ADV601LC during decode. The result is
high quality mixing of different images. This also provides the
capability to fade from one image to the next.
Edge or Motion Detection
In certain remote video surveillance and machine vision applica-
tions, it is desirable to detect edges or motion. Edges can be
quickly found through evaluation of the high frequency blocks.
Motion searches can be achieved in two ways:
• Evaluation of the smallest luminance block. Because the size
• Polling the Sum of Squares registers. Because large changes in
The compressed video bit stream was created with simple
parsing in mind. This type of parsing means that a lower
resolution/lower bandwidth bit stream can be extracted with
little computational burden. Generally, this effect is accom-
plished by selecting a subset of lower frequency blocks. This
technique is useful in applications where the same video
source material must be sent over a range of different commu-
nication pipes {i.e., ISDN (128 Kbps), T1 (1.5 Mbps) or T3
(45 Mbps)}.
In this case, a host CPU could encode a smaller image size
and fill in high frequency blocks with zeros. Again, image
quality would depend on the performance of the host. The
Bin Width may be set to zero, zeroing out the data in any
particular Mallat block.
or decode
Trade off image softness for higher compression. Attenua-
tion of the higher frequency blocks during encode leads to
softer images, but it can lead to much higher compression
performance.
This effect is achieved by controlling gain of low pass chromi-
nance blocks during encode or decode.
This effect is achieved by controlling the gain of the low fre-
quency luminance blocks during encode or decode.
This effect achieved by attenuation of luminance blocks.
of the smallest block is so mcuh smaller than the others, the
computational burden is significantly less than doing an
evaluation over the entire image.
the video data create patterns, it is possible to detect motion
in the video by polling the Sum of Squares registers, looking
for patterns and changes.
REV. 0