I want to make a Raspi based networkcamera but WITHOUT the original raspi Camera. I have a camera that gives me single frames which I store in a 2d array. 1 Byte/Pixel because it is grayscale.
What are the steps I have to do to convert it to a Stream that I can watch via HTTP request or Socket Stream or rtsp.
How do I need to convert the single Frames into something that I can stream via Network?
Do I take a single frame and convert it to jpeg (or mjpeg) or something different?
Programming language is C or C++ nothing else.
Thanks in Advance
Related
I've written a C++ converter based on FFMpeg which can receive a link to hls-stream and convert it into local .mp4 video. So far, so good, converter works like a charm, no questions about that.
PROBLEM: No matter what input source I'm providing to the converter, at the end of convertation I need to receive video with key-frames ONLY. I need such video due to perfect seeking forward and reverse.
It's a well-known fact that subsidiary video frames (P and B) dependent on their owner-frame (I frame), because this frame contains full pixel map. According to that, we can recreate a I frame for each P and B frame by merging their data with their I frame. That's why such ffmpeg command ffmpeg -i video.mp4 output%4d.jpg works.
QUESTION: How can I implement an algorithm of merging of frames in order to recreate Key-frames ONLY video at the end? What kind of quirks I need to know about merging datas of AVPackets?
Thanks.
You cannot "merge" P and B-frames of a compressed stream (e.g. with H.264 codec), to obtain I-frames.
What ffmpeg does with
ffmpeg -i video.mp4 output%4d.jpg
is decoding each frame (thus it needs to start from an I-frame, then decode all subsequent P and B-frames in the stream), and compress them back to JPEG and output a JPEG image for each frame in the original input stream.
If you want to convert an input stream with P/B frames to an intra-only stream (with all I-frames), you need to transcode the stream.
That means decode all frames from the original stream and encode them back to an intra-only stream.
I have an OpenGL application of which rendered images need to be streamed over internet to mobile clients. Previously, it sufficed to simply record the rendering into a video file, which is already working, and now this should be extended to subsequent streaming.
What is working right now:
Render a scene to an OpenGL framebuffer object
Capture the FBO content using NvIFR
Encode it to H.264 using NvENC (no CPU round trip required)
Download the encoded frame to host memory as a byte array
Append this frame to a video file
None of this steps involves FFmpeg or any other library so far. I now want to replace the last step with "Stream the current frame's byte array over internet" and I assume that using FFmpeg and FFserver would be a reasonable choice for this. Am I correct? If not, what would be the proper way?
If so, how do I approach this within my C++ code? As pointed out, the frame is already encoded. Also, there is no sound or other stuff, simply a H.264 encoded frame as byte array that is updated irregularly and should be converted into a steady video stream. I assume that this would be FFmpeg's job and that the subsequent streaming via FFserver would be simple from there. What I don't know is how to feed my data to FFmpeg in the first place, as all FFmpeg tutorials I found (in a non-exhaustive search) work on a file or webcam/capture device as data source, not volatile data in main memory.
The file mentioned above that I am already able to create is a C++ file stream to which I append each single frame, meaning that different framerates of video and rendering are not treated correctly. This also needs to be taken care of at some point.
Can somebody point me in the right direction? Can I forward data from my application to FFmpeg to build a proper video feed without writing to the hard disk? Tutorials are greatly appreciated. By the way FFmpeg/FFserver is not mandatory. If you have a better idea for streaming of OpenGL framebuffer contents, I'm eager to know.
You can feed the ffmpeg process readily encoded H.264 data (-f h264) and tell it to simply copy the stream into to the output multiplexer (-c:v copy). To get the data actually into ffmpeg just launch it as a child process with a pipe connected to its stdin and specify stdin as reading source
FILE *ffmpeg_in = popen("ffmpeg -i /dev/stdin -f h264 -c copy ...", "w");
you can then write your encoded h264 stream to ffmpeg_in.
I use Live555 h.264 stream client to query the frame packets from an IP camera, I use ffmpeg to decode the buffer and analysis the frame by OpenCV.(those pipeline are based on testRTSPClient sample, I decode the h.264 frame buffer in DummySink::afterGettingFrame() by ffmpeg)
And now I wanna stream the frame to another client(remote client) OnDemand mode in real-time, the frame may added the analysis result(boundingboxs, text, etc), how to use Live555 to achieve this?
Well, your best bet is to re-encode the resultant frame (with bounding boxes etc), and pass this to an RTSPServer process which will allow you to connect to it using an rtsp url, and stream the encoded data to any compatible rtsp client. There is a good reference on the FAQ for how to do this http://www.live555.com/liveMedia/faq.html#liveInput which walks you through the steps taken, and provides example source code which you can modify for your needs.
Can I get compressed data from cvCapture, which is stored into buffer after cvGrabFrame, before using cvRetrieveFrame? I need to keep them in the an archive of compressed frames.
I use cvCaptureFromFile for getting rtsp stream from Axis IP camera.
As far as I see it it's not possible. CvCapture is a black box structure. There is not really a possibility to access the data inside.
There is the function cv::imencode to compress images. It is not optimal to compress the image again but it should work.
By the way, you're using the old C API and there is a newer API for C++ that makes capturing more elegant: cv::VideoCapture.
What is the best way to capture streamed MJPEG from a network IP camera?
I'd like to get frames and process them, using c++ (or python extended with c++).
Is OpenCV my best option?
Appart from OpenCV, you can use mplayer with -vo yuv4mpeg redirected to a pipe to get a stream of uncompressed yuv images. You can create the mplayer process and pipe from C++.
Another way is to use a RTSP library (your IP camera probably uses it as protocol)