Setting size and subtype is ignored / Hints for a solution

[bambusch]

Hi Theo,

first of all, thank you for that useful piece of code!

When I tried to set another size and another fourcc, it was always ignored and the default settings were taken.

After a while I figured out, that in the method "setSizeAndSubtype" (Line 1656) you calculate lSampleSize = attemptWidth_attemptHeight_3 (Line 1677).

It always resulted in E_FAIL when calling SetFormat (Line 1685).

The reason is that my Microsoft LifeCam Studio (only?) uses YUY2 which is 2 bytes per pixel (instead of 3). So the lSampleSize must be calculated by attemptWidth_attemptHeight_2 in this case - then it works.

A solution might be an additional array with bytes-per-pixel information for each mediatype.

Regards
Stefan

(Environment: Windows 8 / Visual Studio 2012 / x64)

[ofTheo]

that could be because we only do RGB output.
we're you trying to get YUV2 output?

VI can handle a range of YUV inputs but it converts them to RGB before the pixels get sent back to the user.
( hence why w_h_3 was hardcoded )

[bambusch]

I've just realized, that the code in the opencv source differs in some parts from your original.
(The opencv code allows to setup the device with a fourcc-code.)
Perhaps the described problem only resides in the opencv version.

However, since the method setSizeAndSubtype looks the same in both versions, I will describe my observation:

The method setSizeAndSubtype first retrieves the current format (tmpType). This always resulted in S_OK. But later, when setting the format (VD->pAmMediaType) inside the same method it always resulted in E_FAIL - also when trying all mediaSubtypes in the for-loop (1813 to 1823 in your code).

So I compared the content of the retrieved tmpType and the newly created VD->pAmMediaType and found out, that the lSampleSize of the tmpType only was w_h_2. I changed the calculation of the lSampleSize of VD->pAmMediaType to w_h_2 and it works for me.

This led to my assumption, that the camera itself allows only a two-byte-per-pixel format (YUY2).

(The output image was and is still converted to RGB.)

[ofTheo]

Also got another report of this in #12 - Not sure if you have a patch that figures out the Bytes Per Pixels based on media types.

If you do please send a pull request - otherwise I'll put this on my todo list.