PDA

View Full Version : ML video processing



ehsan2004
09-16-2004, 01:20 PM
hello
some video devices have own properties that limit them.As an example, divo is a totally none-square pixel board. is there limitations for OpenML on these boards?Can we, as an example, capture square pixels on the 'divo'?

[ September 16, 2004: Message edited by: ehsan_kamrani ]

frankb
10-04-2004, 02:13 PM
Are you referring to the SGI divo video I/O board?

ehsan2004
10-04-2004, 08:41 PM
Hi Frank
The Information i have used is from the *Lurker's Guide to video*.so i think that it's the SGI divo I/O board .But actually i want to see that is OpenML individual from hardware?--divo was only an example.
In general, if the hardware doesn't support a capability, can openML support that capability in a software manner?

[ October 05, 2004: Message edited by: ehsan_kamrani ]

frankb
10-11-2004, 02:51 PM
OpenML is intendend for realtime applications. OpenML will not prevent you from supporting a computationally intensive operation/feature in SW but it's not recommended because the design will not scale.
You might be able to get away with one instance of the mlmodule with complex SW video processing (e.g. aspect ratio conv) but as soon as an application adds more devices the cpu will not be able to support the operations in realtime. You will have a better chance of processing lower bandwidth signals like audio with SW in a module.
Ideally a module should setup the HW device, transfer buffers and notify on errors.

ehsan2004
11-13-2004, 12:21 PM
Hi
1)Which type of video card do you suggest me to buy? can i use from the simple capture cards for my OpenML programs?

2)As i know, DirectX consists of two layers: HAL-Hardware abstraction layer- and HEL- Hardware emulation layer. Are there these features(HAL and HEL ) in OpenML?

[ November 15, 2004: Message edited by: ehsan_kamrani ]

fjaubert
11-23-2004, 04:45 AM
Hello Ehsan,

1) for exploratory work and preliminary testing purposes, you can use pretty much any capture card that works with Windows -- the ML SDK comes with a sample module that simply uses the Windows drivers.

Note however that you will not get real-time performance from this setup, since the ML module is not really a driver, it is simply an app-level emulation. So for final integration and performance tests, you need to find a card that has a native ML module. I know of DVS and SGI that have such cards, there may be others as well.

2) No, the ML spec does not have explicit HAL and HEL layers. Remember that ML is only an API, it isn't an implementation (as opposed to DirectX which is both an API *and* an implementation).

The ML reference implementation (ie: the ML SDK) does not use HAL and HEL either. Instead, it has a device-independent layer and a device-dependent layer. But in any case, this is not really relevant to programmers using the API (it *is* relevant to programmers wanting to develop new ML modules).

ehsan2004
11-30-2004, 11:05 AM
Hi Fabrice
For 525-line analog signals, the picture should be produced in this manner: ( F1 has 243 active lines, F2 has 243 active lines = 486 active lines).
So if we want to show each field in the video monitor that has more than 486 lines, what should we do with the other lines of the video monitor?

[ December 04, 2004: Message edited by: ehsan_kamrani ]

fjaubert
12-08-2004, 11:10 AM
We're talking 'video monitor' here, not computer monitor. If your video monitor is NTSC (525-line), it will have 486 active lines, no more. If your video monitor is PAL (625-line), it will have 576 active lines, no more, no less.

ehsan2004
04-18-2005, 07:23 AM
Hi Fabrice
1.Although i have studied the lurker book, but i'm a bit confused about the following terms:
-Interlaced signal
-Interleaved signal
-none-interleaved signal
-Interlaced image
-progressive buffers
-interleaved buffers
Can you please give me some small definitions about the above terms?

2.As i have seen in the book *A technical introduction to digital video*, NTSC has 483 lines and 644 actice pixels.Lurcker book says that there are many types of NTSC.OpenML uses from the parameter ML_TIMING_525 to set the 525 timing.So which 525 timing is supposed to be used in this timing?

3.Another question is about the horizontal blanking.Does it mean that the video monitors have some inactive pixels? Can we see these pixels in the video mointor--As an example, like a black part?

-Ehsan-

[ April 27, 2005: Message edited by: Ehsan Kamrani ]

skolapper
09-05-2009, 01:25 PM
Hi

I have been looking to HDR process videos similar to photogrpahers processing
images.

Is it possible? what kind of software? etc.

I heard there is a company making the hardware and software for real time HDR video processing but I lost track of it.

Regards

Caracalla