Skip to main content
typos
Source Link
Blup1980
  • 6.7k
  • 3
  • 29
  • 49

Your bottleneck in term of processing power will be the live video processing and real time compression. Video compression done in software takes A LOT of computing time.

Simple evaluation:

you use a low res VGA sensor : 640 x 480 pixels black and white.

you have 640 x 480 = 307200 pixels.

you didn't specify the framerate. But let's decide 25 fps for the computation.

Now you have to process 307200 * 25 = 7.68 Megapixel/sec ! or 0.13us per pixel.

Imagine that you have a high-end ARM cortex-M3 microcontroller at 100 MIPS. or 0.01 us/cycle.

TheyThen you have 0.13us / 0.01 = 13 single instructions available per pixel !

This assumeassumes that you CPU is doing nothing else. Which is not your case. Thus your compression algorithm has to be very simple. Or you should find a chip that is able to do this for you in hardware or reduce a lot the frame rate.

Your bottleneck in term of processing power will be the live video processing and real time compression. Video compression done in software takes A LOT of computing time.

Simple evaluation:

you use a low res VGA sensor : 640 x 480 pixels black and white.

you have 640 x 480 = 307200 pixels.

you didn't specify the framerate. But let's decide 25 fps for the computation.

Now you have to process 307200 * 25 = 7.68 Megapixel/sec ! or 0.13us per pixel.

Imagine that you have a high-end ARM cortex-M3 microcontroller at 100 MIPS. or 0.01 us/cycle.

They you have 0.13us / 0.01 = 13 single instructions available per pixel !

This assume that you CPU is doing nothing else. Which is not your case. Thus your compression algorithm has to be very simple. Or you should find a chip that is able to do this for you in hardware or reduce a lot the frame rate.

Your bottleneck in term of processing power will be the live video processing and real time compression. Video compression done in software takes A LOT of computing time.

Simple evaluation:

you use a low res VGA sensor : 640 x 480 pixels black and white.

you have 640 x 480 = 307200 pixels.

you didn't specify the framerate. But let's decide 25 fps for the computation.

Now you have to process 307200 * 25 = 7.68 Megapixel/sec ! or 0.13us per pixel.

Imagine that you have a high-end ARM cortex-M3 microcontroller at 100 MIPS. or 0.01 us/cycle.

Then you have 0.13us / 0.01 = 13 single instructions available per pixel !

This assumes that you CPU is doing nothing else. Which is not your case. Thus your compression algorithm has to be very simple. Or you should find a chip that is able to do this for you in hardware or reduce a lot the frame rate.

Source Link
Blup1980
  • 6.7k
  • 3
  • 29
  • 49

Your bottleneck in term of processing power will be the live video processing and real time compression. Video compression done in software takes A LOT of computing time.

Simple evaluation:

you use a low res VGA sensor : 640 x 480 pixels black and white.

you have 640 x 480 = 307200 pixels.

you didn't specify the framerate. But let's decide 25 fps for the computation.

Now you have to process 307200 * 25 = 7.68 Megapixel/sec ! or 0.13us per pixel.

Imagine that you have a high-end ARM cortex-M3 microcontroller at 100 MIPS. or 0.01 us/cycle.

They you have 0.13us / 0.01 = 13 single instructions available per pixel !

This assume that you CPU is doing nothing else. Which is not your case. Thus your compression algorithm has to be very simple. Or you should find a chip that is able to do this for you in hardware or reduce a lot the frame rate.