hi,

after reading about the performance issues with processings video library (well, quicktime for java seems to be blamed for that) i decided to write a my own motiondetector using jmf.
which works great with 24bit rgb data (where each color has its own byte).
i'm having big trouble trying to convert 16 bit color data to 3 individual rgb-channels (and back) can anyone point me to an example or hint how to do that?

this is what i came up with so far ...
Code:

  
		int r,g,b,c;
		c |= inData[i] & 0xFF - Byte.MIN_VALUE;
		c <<= 8;
		c |= inData[i - 1] & 0xFF - Byte.MIN_VALUE;

		r = (c & 0xF800) >> 11;
		g = (c & 0x07E0) >> 5;
		b = c & 0x1F;
 

is this correct or am i doing something wrong when trying to build the simple greyscale value and inserting it back into the buffer?

Code:

		int bw;
		bw = (int) ((r + g + b) / (double) 3.0);
		outData[i - 1] = (byte) (bw & 0xFF + Byte.MIN_VALUE);
		bw >>= 8;
		outData[i] = (byte) (bw & 0xFF + Byte.MIN_VALUE);
 

thanks a lot!
horst

after many hours staring at the binary code representation of these color values, i found a solution that works

Code:

// 1. convert your byte data to an int
byte[] data; // some byte array e.g. a JMF buffer

// endianness determines which one is the high or low byte
// little endian here:
int highbyte = data [x];
int lowbyte = data [x+1];
int rgb16 = ((highbyte & 0xff) << 8) | (lowbyte & 0xff);

// 2. split that int into 3 ints representing 3 color channels

int red = rgb16 >> 7 & 0xff;
int green = rgb16 >> 2 & 0xff;
int blue = rgb16 << 3 & 0xff;

// 3. merge these 3 channels back into one int
int result = (red >> 3) << 10 | (green >> 3) << 5 | blue >> 3;

// 4. convert the resulting int back to two bytes
byte highbyte = (byte) (number & 0xFF);
byte lowbyte = (byte) (number >> 8 & 0xFF);
 

this was quite tricky to figure out for me, so maybe this saves someone from a headache or two