Design Article
Video-image processing with customizable DSP-FPGA platform
David Rice<br>Vice President and Technical Director,<br>Critical Link
11/7/2008 2:00 AM EST
When it comes to deciding how to handle video/image processing, engineers have a number of choices.
One option is going the ASIC route. In an ideal, money-is-no-object world, this would seem to represents the best option, since it pretty much guarantees a perfect fit. But with an ASIC, you can find yourselves locked into a situation in which yesterday's perfect fit becomes not-so-perfect well before development costs are recouped.
Of course, money always matters. And precisely because of the engineering and tooling costs, the ASIC approach makes sense only when dealing with high-volume applications like HDTVs, digital cameras, cell phones, and MP3 players. For low volume applications, however, the upfront costs can quickly and easily overshadow the benefits.
Building a custom processing system using off-the-shelf processors and FPGAs is another option. However, developing such a system -- even one that's not as fully from scratch as an ASIC would be - can also be quite expensive, with much of the time and effort spent building the infrastructure -- Ethernet and USB interfaces, data acquisition interfaces, etc. -- required to support the applications.
On the other end of the spectrum, a fully off-the-shelf solution that combines an image capture card with a PC to do the image processing may do the trick in some situations. Application examples in which this approach works well include many industrial imaging applications and some security applications. However, imaging technology can be expensive in its own right, so this approach may not always work from a cost angle. More importantly, for embedded applications, having a PC off to the side makes no sense whatsoever.
Embedded systems can, of course, be built using off-the-shelf processing boards and image capture boards. These solutions can save development time over custom approaches, but in many cases, the recurring cost-per-unit can lead to final system costs that exceed what the customer is willing to pay.
With both ASIC and off-the-shelf approaches, you can also get locked in. Once you have an ASIC solution in place, it's expensive to re-engineer and re-tool it. And a purely off-the-shelf solution can put you at the mercy of the marketplace -- e.g., if you design around a specific PC and video card and either becomes obsolete, you are, again, faced with a redesign. The custom processor system approach is better in this regard, if the development costs are tolerable.
A fourth alternative that's highly appropriate for lower volume applications is the use of a CPU platform that combines standard components that have been fine-tuned to support specific application requirements. This approach enables more rapid product development than does an ASIC or custom processor, as well as far greater flexibility than can be achieved by using strictly off-the-shelf components.
Next: MityDSP
One option is going the ASIC route. In an ideal, money-is-no-object world, this would seem to represents the best option, since it pretty much guarantees a perfect fit. But with an ASIC, you can find yourselves locked into a situation in which yesterday's perfect fit becomes not-so-perfect well before development costs are recouped.
Of course, money always matters. And precisely because of the engineering and tooling costs, the ASIC approach makes sense only when dealing with high-volume applications like HDTVs, digital cameras, cell phones, and MP3 players. For low volume applications, however, the upfront costs can quickly and easily overshadow the benefits.
Building a custom processing system using off-the-shelf processors and FPGAs is another option. However, developing such a system -- even one that's not as fully from scratch as an ASIC would be - can also be quite expensive, with much of the time and effort spent building the infrastructure -- Ethernet and USB interfaces, data acquisition interfaces, etc. -- required to support the applications.
On the other end of the spectrum, a fully off-the-shelf solution that combines an image capture card with a PC to do the image processing may do the trick in some situations. Application examples in which this approach works well include many industrial imaging applications and some security applications. However, imaging technology can be expensive in its own right, so this approach may not always work from a cost angle. More importantly, for embedded applications, having a PC off to the side makes no sense whatsoever.
Embedded systems can, of course, be built using off-the-shelf processing boards and image capture boards. These solutions can save development time over custom approaches, but in many cases, the recurring cost-per-unit can lead to final system costs that exceed what the customer is willing to pay.
With both ASIC and off-the-shelf approaches, you can also get locked in. Once you have an ASIC solution in place, it's expensive to re-engineer and re-tool it. And a purely off-the-shelf solution can put you at the mercy of the marketplace -- e.g., if you design around a specific PC and video card and either becomes obsolete, you are, again, faced with a redesign. The custom processor system approach is better in this regard, if the development costs are tolerable.
A fourth alternative that's highly appropriate for lower volume applications is the use of a CPU platform that combines standard components that have been fine-tuned to support specific application requirements. This approach enables more rapid product development than does an ASIC or custom processor, as well as far greater flexibility than can be achieved by using strictly off-the-shelf components.
Next: MityDSP
|
|
|
|
|
Navigate to related information
