Develop Reference

Android MediaCodec realtime h264 encoding/decoding latency

I'm working with Android MediaCodec and use it for a realtime H264 encoding and decoding frames from camera. I use MediaCodec in synchronous manner and render the output to the Surface of decoder and everething works fine except that I have a long latency from a realtime, it takes 1.5-2 seconds and I'm very confused why is it so.
I measured a total time of encoding and decoding processes and it keeps around 50-65 milliseconds so I think the problem isn't in them.
I tried to change the configuration of the encoder but it didn't help and currently it configured like this:
val formatEncoder = MediaFormat.createVideoFormat("video/avc", 1920, 1080)
formatEncoder.setInteger(MediaFormat.KEY_FRAME_RATE, 30)
formatEncoder.setInteger(MediaFormat.KEY_I_FRAME_INTERVAL, 5)
formatEncoder.setInteger(MediaFormat.KEY_BIT_RATE, 1920 * 1080)
formatEncoder.setInteger(MediaFormat.KEY_COLOR_FORMAT, MediaCodecInfo.CodecCapabilities.COLOR_FormatSurface)
val encoder = MediaCodec.createEncoderByType("video/avc")
encoder.configure(formatEncoder, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE)
val inputSurface = encoder.createInputSurface() // I use it to send frames from camera to encoder
encoder.start()
Changing the configuration of the decoder also didn't help me at all and currently I configured it like this:
val formatDecoder = MediaFormat.createVideoFormat("video/avc", 1920, 1080)
val decoder = MediaCodec.createDecoderByType("video/avc")
decoder.configure(formatDecoder , outputSurface, null, 0) // I use outputSurface to render decoded frames into it
decoder.start()
I use the following timeouts for waiting for available encoder/decoder buffers I tried to reduce their values but it didn't help me and I left them like this:
var TIMEOUT_IN_BUFFER = 10000L // microseconds
var TIMEOUT_OUT_BUFFER = 10000L // microseconds
Also I measured the time of consuming the inputSurface a frame and this time takes 0.03-0.05 milliseconds so it isn't a bottleneck. Actually I measured all the places where a bottleneck could be, but I wasn't found anything and I think the problem is in the encoder or decoder itself or in their configurations, or maybe I should use some special routine for sending frames to encoding/decoding..
I also tried to use HW accelerated codec and it's the only thing that helped me, when I use it the latency reduces to ~ 500-800 milliseconds but it still doesn't fit me for a realtime streaming.
It seems to me that the encoder or decoder buffers several frames before start displaying them on the surface and eventually it leads to the latency and if it really so then how can I disable bufferization or reduce the time of it?
Please help me I'm stucking on this problem for about half a year and have no idea how to reduce the latency, I'm sure that it's possible because popular apps like Telegram, Viber, WhatsApp etc. work fine and without latency so what's the secret here?
UPD 07.07.2021:
I still haven't found a solution to get rid of the latency. I've tried to change h264 profiles, increase and decrease I-frame inteval, bitrate, framerate, but result the same, the only thing that hepls a little to reduce the latency - downgrade the resolution from 1920x1080 to e.g. 640x480, but this "solution" doesn't suit me because I want to encode/decode a realtime video with 1920x1080 resolution.
UPD 08.07.2021:
I found out that if I change the values of TIMEOUT_IN_BUFFER and TIMEOUT_OUT_BUFFER from 10_000L to 100_000L it decreases the latency a bit but increases the delay of showing the first frame quite a lot after start encoding/decoding process.

It's possible your encoder is producing B frames -- bilinear interpolation frames. They increase quality and latency, and are great for movies. But no good for low-latency applications.
Key frames = I (interframes)
Predicted frames = P (difference from previous frames)
Interpolated frames = B
A sequence of frames including B frames might look like this:
IBBBPBBBPBBBPBBBI
11111111
12345678901234567
The encoder must encode each P frame, and the decoder must decode it, before the preceding B frames make any sense. So in this example the frames get encoded out of order like this:
1 5 2 3 4 9 6 7 8 13 10 11 12 17 17 13 14 15
In this example the decoder can't handle frame 2 until the encoder has sent frame 5.
On the other hand, this sequence without B frames allows coding and decoding the frames in order.
IPPPPPPPPPPIPPPPPPPPP
Try using the Constrained Baseline Profile setting. It's designed for low latency and low power use. It suppresses B frames. I think this works.
mediaFormat.setInteger(
"profile",
CodecProfileLevel.AVCProfileConstrainedBaseline);

I believe android h264 decoder have latency (at-least in most cases i've tried). Probably that's why android developers added PARAMETER_KEY_LOW_LATENCY from API level 30.
However I could decrease the delay some frames by querying for the output some more times.
Reason: no idea. It's just result of boring trial and errors
int inputIndex = m_codec.dequeueInputBuffer(-1);// Pass in -1 here bc we don't have a playback time reference
if (inputIndex >= 0) {
ByteBuffer buffer;
if (android.os.Build.VERSION.SDK_INT >= android.os.Build.VERSION_CODES.LOLLIPOP) {
buffer = m_codec.getInputBuffer(inputIndex);
} else {
ByteBuffer[] bbuf = m_codec.getInputBuffers();
buffer = bbuf[inputIndex];
}
buffer.put(frame);
// tell the decoder to process the frame
m_codec.queueInputBuffer(inputIndex, 0, frame.length, 0, 0);
}
MediaCodec.BufferInfo info = new MediaCodec.BufferInfo();
int outputIndex = m_codec.dequeueOutputBuffer(info, 0);
if (outputIndex >= 0) {
m_codec.releaseOutputBuffer(outputIndex, true);
}
outputIndex = m_codec.dequeueOutputBuffer(info, 0);
if (outputIndex >= 0) {
m_codec.releaseOutputBuffer(outputIndex, true);
}
outputIndex = m_codec.dequeueOutputBuffer(info, 0);
if (outputIndex >= 0) {
m_codec.releaseOutputBuffer(outputIndex, true);
}

You need to configure customized(or KEY_LOW_LATENCY if it is supported) low latency parameters for different cpu venders. It is a common problem for android phone.
Check this code https://github.com/moonlight-stream/moonlight-android/blob/master/app/src/main/java/com/limelight/binding/video/MediaCodecHelper.java

Android camera2 jpeg framerate

I am trying to save image sequences with fixed framerates (preferably up to 30) on an android device with FULL capability for camera2 (Galaxy S7), but I am unable to a) get a steady framerate, b) reach even 20fps (with jpeg encoding). I already included the suggestions from Android camera2 capture burst is too slow.
The minimum frame duration for JPEG is 33.33 milliseconds (for resolutions below 1920x1080) according to
characteristics.get(CameraCharacteristics.SCALER_STREAM_CONFIGURATION_MAP).getOutputMinFrameDuration(ImageFormat.JPEG, size);
and the stallduration is 0ms for every size (similar for YUV_420_888).
My capture builder looks as follows:
captureBuilder.set(CaptureRequest.CONTROL_AE_MODE, CONTROL_AE_MODE_OFF);
captureBuilder.set(CaptureRequest.SENSOR_EXPOSURE_TIME, _exp_time);
captureBuilder.set(CaptureRequest.CONTROL_AE_LOCK, true);
captureBuilder.set(CaptureRequest.SENSOR_SENSITIVITY, _iso_value);
captureBuilder.set(CaptureRequest.LENS_FOCUS_DISTANCE, _foc_dist);
captureBuilder.set(CaptureRequest.CONTROL_AF_MODE, CONTROL_AF_MODE_OFF);
captureBuilder.set(CaptureRequest.CONTROL_AWB_MODE, _wb_value);
// https://stackoverflow.com/questions/29265126/android-camera2-capture-burst-is-too-slow
captureBuilder.set(CaptureRequest.EDGE_MODE,CaptureRequest.EDGE_MODE_OFF);
captureBuilder.set(CaptureRequest.COLOR_CORRECTION_ABERRATION_MODE, CaptureRequest.COLOR_CORRECTION_ABERRATION_MODE_OFF);
captureBuilder.set(CaptureRequest.NOISE_REDUCTION_MODE, CaptureRequest.NOISE_REDUCTION_MODE_OFF);
captureBuilder.set(CaptureRequest.CONTROL_AF_TRIGGER, CaptureRequest.CONTROL_AF_TRIGGER_CANCEL);
// Orientation
int rotation = getWindowManager().getDefaultDisplay().getRotation();
captureBuilder.set(CaptureRequest.JPEG_ORIENTATION,ORIENTATIONS.get(rotation));
Focus distance is set to 0.0 (inf), iso is set to 100, exposure-time 5ms. Whitebalance can be set to OFF/AUTO/ANY VALUE, it does not impact the times below.
I start the capture session with the following command:
session.setRepeatingRequest(_capReq.build(), captureListener, mBackgroundHandler);
Note: It does not make a difference if I request RepeatingRequest or RepeatingBurst..
In the preview (only texture surface attached), everything is at 30fps.
However, as soon as I attach an image reader (listener running on HandlerThread) which I instantiate like follows (without saving, only measuring time between frames):
reader = ImageReader.newInstance(_img_width, _img_height, ImageFormat.JPEG, 2);
reader.setOnImageAvailableListener(readerListener, mBackgroundHandler);
With time-measuring code:
ImageReader.OnImageAvailableListener readerListener = new ImageReader.OnImageAvailableListener() {
#Override
public void onImageAvailable(ImageReader myreader) {
Image image = null;
image = myreader.acquireNextImage();
if (image == null) {
return;
}
long curr = image.getTimestamp();
Log.d("curr- _last_ts", "" + ((curr - last_ts) / 1000000) + " ms");
last_ts = curr;
image.close();
}
}
I get periodically repeating time differences like this:
99 ms - 66 ms - 66 ms - 99 ms - 66 ms - 66 ms ...
I do not understand why these take double or triple the time that the stream configuration map advertised for jpeg? The exposure time is well below the frame duration of 33ms. Is there some other internal processing happening that I am not aware of?
I tried the same for the YUV_420_888 format, which resulted in constant time-differences of 33ms. The problem I have here is that the cellphone lacks the bandwidth to store the images fast enough (I tried the method described in How to save a YUV_420_888 image?). If you know of any method to compress or encode these images fast enough myself, please let me know.
Edit: From the documentation of getOutputStallDuration: "In other words, using a repeating YUV request would result in a steady frame rate (let's say it's 30 FPS). If a single JPEG request is submitted periodically, the frame rate will stay at 30 FPS (as long as we wait for the previous JPEG to return each time). If we try to submit a repeating YUV + JPEG request, then the frame rate will drop from 30 FPS." Does this imply that I need to periodically request a single capture()?
Edit2: From https://developer.android.com/reference/android/hardware/camera2/CaptureRequest.html: "The necessary information for the application, given the model above, is provided via the android.scaler.streamConfigurationMap field using getOutputMinFrameDuration(int, Size). These are used to determine the maximum frame rate / minimum frame duration that is possible for a given stream configuration.
Specifically, the application can use the following rules to determine the minimum frame duration it can request from the camera device:
Let the set of currently configured input/output streams be called S.
Find the minimum frame durations for each stream in S, by looking it up in android.scaler.streamConfigurationMap using getOutputMinFrameDuration(int, Size) (with its respective size/format). Let this set of frame durations be called F.
For any given request R, the minimum frame duration allowed for R is the maximum out of all values in F. Let the streams used in R be called S_r.
If none of the streams in S_r have a stall time (listed in getOutputStallDuration(int, Size) using its respective size/format), then the frame duration in F determines the steady state frame rate that the application will get if it uses R as a repeating request."

The JPEG output is by way not the fastest way to fetch frames. You can accomplish this a lot faster by drawing the frames directly onto a Quad using OpenGL.
For burst capture, a faster solution would be capturing the images to RAM without encoding them, then encoding and saving them asynchronously.
On this website you can find a lot of excellent code related to android multimedia in general.
This specific program uses OpenGL to fetch the pixel data from an MPEG video. It's not difficult to use the camera as input instead of a video. You can basically use the texture used in the CodecOutputSurface class from the mentioned program as output texture for your capture request.

A possible solution I found consists of using and dumping YUV without encoding it as JPEG in combination with a micro Sd-card that is able to save up to 95Mb per second. (I had the misconception that YUV images would be larger, so with a cellphone that has full support for the camera2-pipeline, the write speed should be the limiting factor.
With this setup, I was able to achieve the following stable rates:
1920x1080, 15fps (approx. 4Mb * 15 == 60Mb/sec)
960x720, 30fps. (approx. 1.5Mb * 30 == 45Mb/sec)
I then encode the images offline from YUV to PNG using a python script.

Any acoustic echo cancellation (AEC) library capable of 48 kHz?

I'm developing a VoIP application that runs at the sampling rate of 48 kHz. Since it uses Opus, which uses 48 kHz internally, as its codec, and most current Android hardware natively runs at 48 kHz, AEC is the only piece of the puzzle I'm missing now. I've already found the WebRTC implementation but I can't seem to figure out how to make it work. It looks like it corrupts the memory randomly and crashes the whole thing sooner or later. When it doesn't crash, the sound is kinda chunky as if it's quieter for the half of the frame. Here's my code that processes a 20 ms frame:
webrtc::SplittingFilter* splittingFilter;
webrtc::IFChannelBuffer* bufferIn;
webrtc::IFChannelBuffer* bufferOut;
webrtc::IFChannelBuffer* bufferOut2;
// ...
splittingFilter=new webrtc::SplittingFilter(1, 3, 960);
bufferIn=new webrtc::IFChannelBuffer(960, 1, 1);
bufferOut=new webrtc::IFChannelBuffer(960, 1, 3);
bufferOut2=new webrtc::IFChannelBuffer(960, 1, 3);
// ...
int16_t* samples=(int16_t*)data;
float* fsamples[3];
float* foutput[3];
int i;
float* fbuf=bufferIn->fbuf()->bands(0)[0];
// convert the data from 16-bit PCM into float
for(i=0;i<960;i++){
fbuf[i]=samples[i]/(float)32767;
}
// split it into three "bands" that the AEC needs and for some reason can't do itself
splittingFilter->Analysis(bufferIn, bufferOut);
// split the frame into 6 consecutive 160-sample blocks and perform AEC on them
for(i=0;i<6;i++){
fsamples[0]=&bufferOut->fbuf()->bands(0)[0][160*i];
fsamples[1]=&bufferOut->fbuf()->bands(0)[1][160*i];
fsamples[2]=&bufferOut->fbuf()->bands(0)[2][160*i];
foutput[0]=&bufferOut2->fbuf()->bands(0)[0][160*i];
foutput[1]=&bufferOut2->fbuf()->bands(0)[1][160*i];
foutput[2]=&bufferOut2->fbuf()->bands(0)[2][160*i];
int32_t res=WebRtcAec_Process(aecState, (const float* const*) fsamples, 3, foutput, 160, 20, 0);
}
// put the "bands" back together
splittingFilter->Synthesis(bufferOut2, bufferIn);
// convert the processed data back into 16-bit PCM
for(i=0;i<960;i++){
samples[i]=(int16_t) (CLAMP(fbuf[i], -1, 1)*32767);
}
If I comment out the actual echo cancellation and just do the float conversion and band splitting back and forth, it doesn't corrupt the memory, doesn't sound weird and runs indefinitely. (I do pass the farend/speaker signal into AEC, I just didn't want to make the mess of my code by including it in the question)
I've also tried Android's built-in AEC. While it does work, it upsamples the captured signal from 16 kHz.

Unfortunately, there is no free AEC package that support 48khz. So, either move to 32khz or use a commercial AEC package at 48khz.

Streaming video of android surface

I have a problem with rtmp streaming of android surface to a client application. My solution has a very big latency, because my surface is not producing frames 60 times a second, it can produce it in any time (once in 30 seconds for example). So I want to show each new produced frame to the client immediately.
Android is pushing every frame, it looks fine. Client app (jwplayer or vlc) receives it, but it waiting for something. It becomes showing video only after receiving a number of frames. But I need to see every incoming frame on the client side when it just have been received.
How it is working now:
I have a Surface object, obtained from MediaCodec class. MediaCodec is set for h264 video encoding.
MediaCodec mEncoder;
.....
MediaFormat format = MediaFormat.createVideoFormat("video/avc", width, height);
format.setInteger(MediaFormat.KEY_COLOR_FORMAT, colorFormat);
format.setInteger(MediaFormat.KEY_BIT_RATE, videoBitrate);
format.setInteger(MediaFormat.KEY_FRAME_RATE, videoFramePerSecond);
format.setInteger(MediaFormat.KEY_I_FRAME_INTERVAL, iframeInterval);
try {
mEncoder = MediaCodec.createEncoderByType("video/avc");
} catch (IOException e) {
e.printStackTrace();
}
mEncoder.configure(format, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE);
mSurface = mEncoder.createInputSurface();
if (mSurfaceCallback!=null)
mSurfaceCallback.onSurfaceCreated(mSurface);
mEncoder.start();
Sometimes android is drawing to the surface. I can't control the rate of this drawings. Also I can't draw anything to that surface. When something is changed on the surface, MediaCodec is producing new byteBuffer with h264 frame. I send this frame by rtmp.
On a client side I have html page with jwplayer
<pre id="myElement"></pre>
<script>
var playerInstance = jwplayer("myElement");
playerInstance.setup({
file:"rtmp://127.0.0.1:1935/live/stream",
height: 800,
width: 480,
autostart: true,
controls: false,
rtmp: {
bufferlength: 0.1
}
});
</script>
I've tried to change iframeInterval, fps of encoding, bufferlength.. Nothing is really helpful.
Is there is any possibility to show incomming frames immeditely?

What do you mean?
If I understood right - you have:
vlc(client) ---- rtmp protocol ---- android (producer)
You encode video from something (may be camera) using MediaCodec and in vlc there is time latency? right?
At first - what are you using - direct input buffer or MediaCodec.Callback() ?
In callback - you can check every frame in onOutputBufferAvailable and calculate time from one frame to another - this will show you - is this problem on android side.
Then you can try to resolve frame transef problem
You can use WireShark to determine frame sending timing and chek - may be this is network problem
Than - vlc and other players try to fill some internal buffer and only after this starting to show video. Try to turn of vlc buffer (https://forum.videolan.org/viewtopic.php?t=40408). Then - there is common that vlc waiting for IDR frame. You can set interval for sending IDR frames in code
format.setInteger(MediaFormat.KEY_I_FRAME_INTERVAL, iframeInterval);
iframeInterval in seconds (try to set 1 second)
(this will increase streaming size)
Sorry for my bad english

You can hopefully generate video frames at constant rate, even more than 20 fps to produce smooth video with acceptable latency. h264 encoder will handle a stable picture (one changing once in ~30 sec) gracefully, and when there is no change, frame size will be minimal.

SurfaceTexture's onFrameAvailable() method always called too late

I'm trying to get the following MediaExtractor example to work:
http://bigflake.com/mediacodec/ - ExtractMpegFramesTest.java (requires 4.1, API 16)
The problem I have is that outputSurface.awaitNewImage(); seems to always throw RuntimeException("frame wait timed out"), which is thrown whenever the mFrameSyncObject.wait(TIMEOUT_MS) call times out. No matter what I set TIMEOUT_MS to be, onFrameAvailable() always gets called right after the timeout occurs. I tried with 50ms and with 30000ms and it's the same.
It seems like the onFrameAvailable() call can't be done while the thread is busy, and once the timeout happens which ends the thread code execution, it can parse the onFrameAvailable() call.
Has anyone managed to get this example to work, or knows how MediaExtractor is supposed to work with GL textures?
Edit: tried this on devices with API 4.4 and 4.1.1 and the same happens on both.
Edit 2:
Got it working on 4.4 thanks to fadden. The issue was that the ExtractMpegFramesWrapper.runTest() method called th.join(); which blocked the main thread and prevented the onFrameAvailable() call from being processed. Once I commented th.join(); it works on 4.4. I guess maybe the ExtractMpegFramesWrapper.runTest() itself was supposed to run on yet another thread so the main thread didn't get blocked.
There was also a small issue on 4.1.2 when calling codec.configure(), it gave the error:
A/ACodec(2566): frameworks/av/media/libstagefright/ACodec.cpp:1041 CHECK(def.nBufferSize >= size) failed.
A/libc(2566): Fatal signal 11 (SIGSEGV) at 0xdeadbaad (code=1), thread 2625 (CodecLooper)
Which I solved by adding the following before the call:
format.setInteger(MediaFormat.KEY_MAX_INPUT_SIZE, 0);
However the problem I have now on both 4.1.1 (Galaxy S2 GT-I9100) and 4.1.2 (Samsung Galaxy Tab GT-P3110) is that they both always set info.size to 0 for all frames. Here is the log output:
loop
input buffer not available
no output from decoder available
loop
input buffer not available
no output from decoder available
loop
input buffer not available
no output from decoder available
loop
input buffer not available
no output from decoder available
loop
submitted frame 0 to dec, size=20562
no output from decoder available
loop
submitted frame 1 to dec, size=7193
no output from decoder available
loop
[... skipped 18 lines ...]
submitted frame 8 to dec, size=6531
no output from decoder available
loop
submitted frame 9 to dec, size=5639
decoder output format changed: {height=240, what=1869968451, color-format=19, slice-height=240, crop-left=0, width=320, crop-bottom=239, crop-top=0, mime=video/raw, stride=320, crop-right=319}
loop
submitted frame 10 to dec, size=6272
surface decoder given buffer 0 (size=0)
loop
[... skipped 1211 lines ...]
submitted frame 409 to dec, size=456
surface decoder given buffer 1 (size=0)
loop
sent input EOS
surface decoder given buffer 0 (size=0)
loop
surface decoder given buffer 1 (size=0)
loop
surface decoder given buffer 0 (size=0)
loop
surface decoder given buffer 1 (size=0)
loop
[... skipped 27 lines all with size=0 ...]
surface decoder given buffer 1 (size=0)
loop
surface decoder given buffer 0 (size=0)
output EOS
Saving 0 frames took ? us per frame // edited to avoid division-by-zero error
So no images get saved. However the same code and video works on 4.3. The video I am using is an .mp4 file with "H264 - MPEG-4 AVC (avc1)" video codec and "MPEG AAAC Audio (mp4a)" audio codec.
I also tried other video formats, but they seem to die even sooner on 4.1.x, while both work on 4.3.
Edit 3:
I did as you suggested, and it seems to save the frame images correctly. Thank you.
Regarding KEY_MAX_INPUT_SIZE, I tried not setting, or setting it to 0, 20, 200, ... 200000000, all with the same result of info.size=0.
I am now unable to set the render to a SurfaceView or TextureView on my layout. I tried replacing this line:
mSurfaceTexture = new SurfaceTexture(mTextureRender.getTextureId());
with this, where surfaceTexture is a SurfaceTexture defined in my xml-layout:
mSurfaceTexture = textureView.getSurfaceTexture();
mSurfaceTexture.attachToGLContext(mTextureRender.getTextureId());
but it throws a weird error with getMessage()==null on the second line. I couldn't find any other way to get it to draw on a View of some kind. How can I change the decoder to display the frames on a Surface/SurfaceView/TextureView instead of saving them?

The way SurfaceTexture works makes this a bit tricky to get right.
The docs say the frame-available callback "is called on an arbitrary thread". The SurfaceTexture class has a bit of code that does the following when initializing (line 318):
if (this thread has a looper) {
handle events on this thread
} else if (there's a "main" looper) {
handle events on the main UI thread
} else {
no events for you
}
The frame-available events are delivered to your app through the usual Looper / Handler mechanism. That mechanism is just a message queue, which means the thread needs to be sitting in the Looper event loop waiting for them to arrive. The trouble is, if you're sleeping in awaitNewImage(), you're not watching the Looper queue. So the event arrives, but nobody sees it. Eventually awaitNewImage() times out, and the thread returns to watching the event queue, where it immediately discovers the pending "new frame" message.
So the trick is to make sure that frame-available events arrive on a different thread from the one sitting in awaitNewImage(). In the ExtractMpegFramesTest example, this is done by running the test in a newly-created thread (see the ExtractMpegFramesWrapper class), which does not have a Looper. (For some reason the thread that executes CTS tests has a looper.) The frame-available events arrive on the main UI thread.
Update (for "edit 3"): I'm a bit sad that ignoring the "size" field helped, but pre-4.3 it's hard to predict how devices will behave.
If you just want to display the frame, pass the Surface you get from the SurfaceView or TextureView into the MediaCodec decoder configure() call. Then you don't have to mess with SurfaceTexture at all -- the frames will be displayed as you decode them. See the two "Play video" activities in Grafika for examples.
If you really want to go through a SurfaceTexture, you need to change CodecOutputSurface to render to a window surface rather than a pbuffer. (The off-screen rendering is done so we can use glReadPixels() in a headless test.)