I want to have two aplications simultaneously run: one that analyzes image from webcam written using OpenCV (the image is acquired through callback function) and an application that goes into fullscreen mode (let's say a 3D game). The problem is that while the fullscreen mode is launched the webcam image stream is stopping - the frames simply don't turn up, the callback function isn't called. This seems to be an issue with OpenCV - to test that a simple application displaying the image form camera has been prepared.
Why the image stream could be blocked by the fullscreen mode? How to bypass this?
Thanks for any hints.
Your question does not tell if you have tried to search for the problem in the OpenCV community first, so I post this as a hint in case: http://tech.groups.yahoo.com/group/OpenCV/
Also check out the list of issues, maybe its a known bug: https://code.ros.org/trac/opencv/report/1
I'm not an OpenCV expert so this is closer to a suggestion than an answer - but I've experienced similar on my multi-monitor setup using a number of media players on the second monitor and some fullscreen apps ont he first.
In my limited testing, it comes down to what method is used to render the 3d app - DirectX seems to stop media players, OpenGL doesn't.
So it might not be OpenCV which has a problem - it may be what DirectX does to the hardware during a full-screen game.
Actually the behaviour of the OpenCV camera stream is strange. It seems to depend on the native OpenCV window (cvNamedWindow()) that shows the output image form webcam. If the window is on the same screen that went fullscreen the streaming will continue. If the camera window would be placed on another screen, the stream would stop.
Another curious thing is with the screen resolution change. If you change the resolution of the screen and the camera window is not visible (is closed or even minimalized) the image stream would get blocked.
These are just my observations on the topic, maybe it'll be helpful for someone.
Related
I am trying to capture screenshots for OpenGL based games on Windows. Most answers on the internet are to make the window visible and take screenshots. But when I use OBS(Open Broadcaster Software) to broadcast my game play, it is able to streaming the game content even the game window is minimized.
Can anyone help me understand how they do that?
Thanks!
The OpenGL API provides a way to draw to an internal "Frame buffer object" (FBO) which can be blitted to the window. If OBS doesn't provide a way to access that FBO, then your only solution is to make the window visible. Well, you could also hack the broadcast and try to figure out about those bytes...
EDIT:
When the OP clarified his question I was curious. So I dug into OBS code and found that for Windows it uses .dll injection and process hook.
I also found that it hooks "swap buffers" and get the current Frame Buffer with glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING...), copies it and then let Windows continues its job.
Please, notice that I can be wrong. But what I found looks good to me.
I'm developing a cross-platform photo retouching application based on Gtk-2 (but already able to support Gtk-3 with minor modifications).
In my program, the result of the image retouching is previewed in a scrollable area that is implemented through a Gtk::DrawingArea inserted into a Gtk::ScrolledWindow. The drawing itself is performed using Cairo.
Recently I had the possibility to test the software on a MacBookPro laptop with retina display, and I've immediately realised that the preview image gets magnified by a factor of 2, like all the rest of the GUI elements.
Is there a way to tell Cairo and the DrawingArea to use the native screen resolution, instead of applying a 2x magnification? Is this supported in recent Gtk-3 versions?
Thanks for you help.
Am currently working on a new application at work and am needing some help.
Basically I need to take real time screenshots of an application running on the users desktop(not my application) then warp the image using a shade and display the wrapped screenshot at 60 frames a second.
I have a windows form application that does this but I am getting 6 frames a second and when I scale this up to 1600x1600 it drops to 2fps.
So I thought that if I could do this in xna and move as much as possible to the GPU it would speed it up.
So I have googled and googled and cant find any information on how to get a screenshot of a running application in xna other than my xna application.
Thanks for any help.
You can use XNA or additionaly the DWM or GDI+ to get the full desktop. Here are some links which can help you:
Capture screenshot of hidden desktop
Is there a way to take a screenshot of the user's Windows desktop?
C: take screenshot
What is the best way to take screenshots of a Window with C++ in Windows?
Take screen shot in XNA
I think there is no way to get a window screenshot of another application. You can use some pattern detection code to find out where a window is but I do not recommend this.
Is there any way to get window handle of the window which is currently playing video. This is the only information my program will be having.
updated to include info incorrectly provided as an answer
I think I should explain what exactly I want to achieve here.
I actually wanted to share/stream my DVD data to the remote machine. Currently what I am doing is, I am capturing the screen/video into to still frames and sending it to remote system but now I don't want to see the playing video on my host machine. I could think of few probable solutions,
1.) If we can capture the data of a hide/minimized window.
Did some investigation and seems it is not possible. Please add your thoughts.
2.) Convert the DVD data format into ffmpeg format and stream it.
Don't have any idea if we will be allowed to convert the data format. If most of the DVD formats allow to covert then I can go for this option but not sure how complicated it could be.
3.) Will create some virtual surface play the DVD data to that surface and capture the screen of that surface.
Again not sure if DVD will play on that virtual/fake surface created by kernel mode driver.
There are probably three main playback engines used on windows; DirectShow (WMP, MPC) , ffmpeg (VLC, MPlayer) and QuickTime.
If you look closer at DirectShow will will see that it supports hardware overlays, windowed and windowless rendering and Direct3d surface support.
Even if you focus on a single app you are going to have problems since you don't know what kind of renderer is in use. You might be able to find a child window that always has the same position and dimensions as the video, but then you are relying on things that could change between versions etc.
I have a windows application that scrapes pixels from the screen for recording (in the form of a video) to a custom screen-sharing format. The problem is that on machines using a software cursor, blitting from the screen with SRCCOPY|CAPTUREBLIT (so that layered windows also show up in the image) causes the cursor to blink, as described in Case of the Disappearing Cursor.
For single screen shots, this is not a problem, but when multiple screen shots are taken in rapid succession, the cursor blinks so fast that it sometimes seems to disappear altogether.
I have looked into using the Windows Media Encoder SDK (as described in a codeproject article, see below) because it doesn't cause the cursor to blink, but there seems to be no way to directly access the frame data. Unfortunately, both real-time encoding and the custom format are both requirements, which makes windows Media Encoder unusable for this purpose.
I have also tried the DirectX way (described in the same article, see below), and it seems to suffer from the same problem.
Has anyone else run into this problem? There must be a way around it - many commercial screen sharing programs have no such problem.
article: www.codeproject.com/KB/dialog/screencap.aspx
you can use Magnification API in windows vista or later.
i cannot find a good idea in windows xp.
What about using a mirror driver?
You are right, a mirror would certainly work. However, at the moment, I am trying to stay away from that approach because of the security and permissions concerns when installing under a user without admin rights. Correct me if I am wrong, but I don't think there is any way to install a driver without such rights. Besides that, it seems that that would be needlessly complex: there should be a simpler / less invasive way to do this. (I should have mentioned this in my original question)
Just copy the screen and the cursor separately and overlay them.
The thought I had to overcome the flicker is to "manually" draw "your own copy of the mouse", then make the BitBlt call, or to call BitBlt with just SRCCOPY then manually capture any visible transparent windows over the top of it. I don't know how the commercial stuff does it (or the windows media encoder apparently does).
ref: http://us.generation-nt.com/xp-bitblt-captureblt-option-help-26970632.html