I am trying to setup camera preview through ImageReader(YUV_420_888 format) using Camera2 API. First of all i need to choose supported preview size:
StreamConfigurationMap scmap = camCharacteristics.get(CameraCharacteristics.SCALER_STREAM_CONFIGURATION_MAP);
Size previewSizes[] = scmap.getOutputSizes(ImageReader.class);
My Nexus 5X supports next sizes:
[4032x3024,4000x3000,3840x2160,3288x2480,3200x2400,2592x1944,2688x1512,2048x1536,1920x1080,1600x1200,1440x1080,1280x960,1280x768,1280x720,1024x768,800x600,864x480,800x480,720x480,640x480,640x360,352x288,320x240,176x144,160x120]
Prepare an instance of ImageReader and start CaptureSession with repeating capture request:
mImageReader = ImageReader.newInstance(W,H, ImageFormat.YUV_420_888,1);
Then I'm trying to read each preview frame (for further processing and displaying via GLES) in OnImageAvailableListener. And what i want to know - how many Y-channel bytes i have received:
public void onImageAvailable(ImageReader reader) {
ByteBuffer yBuffer = mImageReader.acquireNextImage().getPlanes()[0].getBuffer();
Log.d("Camera2Debug","Y-channel bytes received: " + yBuffer.remaining());
...
}
The Y-channel for YUV_420_888 image should contain WxH bytes, where W - is the width, and H - is the height of the considered image.
Problem: For some supported preview sizes the actual size of yBuffer doesn't match with expected value (WxH).
For example:
Preview Size | Y-bytes received | Y-bytes expected | match
4032x3024 | 12 192 768 | 12 192 768 | yes
1920x1080 | 2 073 600 | 2 073 600 | yes
1440x1080 | 1 589 728 | 1 555 200 | no
1280x960 | 1 228 800 | 1 228 800 | yes
1280x768 | 983 040 | 983 040 | yes
800x600 | 499 168 | 480 000 | no
...
499168
So due to this problem I can't use the neccessary preview size even if it is supported by device.
What am I doing wrong?
You're likely not taking into account https://developer.android.com/reference/android/media/Image.Plane.html#getRowStride(), which may be larger than the width (but is at least equal to it). The actual buffer size will be
size = rowStride * height - (rowStride - width)
with the subtraction of (rowStride - width) because the last row doesn't include any padding past the last pixel.