CODEWITHSUNDEEP
androidimage-processingcameraandroid-imageandroid-camera2
Sarah Sukardi
Sarah Sukardi

Reputation: 45

Android Camera2 YUV_420_888 Y Channel Interpretation

I am writing a simple Android application where I retrieve frame buffers using ImageReader.OnImageAvailableListener to acquire the latest image and process the Y plane. I use this plane to calculate a simple metric whose value will decide whether I display something on the screen or not.

My question was how to process and interpret only the Y plane (I have no need for the U and the V planes). It is to my understanding that the Y plane contains the luminance channel; ie. the channel with the brightness of the image.

Documentation states that each pixel is 8 bits, ie, one byte, which I guess makes sense since the pixels are in a ByteBuffer. However, after some experiments printing the first byte of the Y channel (I do this by calling Log.i(TAG, String.valueOf(image.getPlanes()[0].getBuffer().get(0) in my ImageReader.OnImageAvailableListener), I don't see any correlation between image "brightness" and the value of the byte. For example, when I put the camera on a completely black background, I get values from ~0 to 15 for the luminance. When I put it on a bright white background, I get values around -110. However, these aren't some kind of minima or maxima; I'll get values that don't seem to correlate with luminance at all when on a non all-white or all-black background.

My question is, how should these luminance values be interpreted, and are they correlated in any way at all with the brightness of the image at a particular pixel? Am I retrieving the luminance values for the frame buffer correctly, and if not, how should one retrieve them?

Upvotes: 2

Views: 612

Answers (1)

Sarah Sukardi
Sarah Sukardi

Reputation: 45

I finally figured it out after a lot of hand-wringing: I stumbled upon the key to the answer on this link. The values that are being reported are in fact the luminance values, but the ByteBuffer in which they are placed is SIGNED, while the luminance values are UNSIGNED (as reported in the link). Thus, each value from the buffer should be &-ed with 0xff to retrieve the true luminance value, which this time makes sense.

Upvotes: 1

Related Questions