Separate RGB layers of a RAW image

I am trying to separate the RGB layers of a RAW image. I'm able to get the red layer using rawr=img(:,:,1), but when I'm trying to get the green and blue layer using the command rawg=img(:,:,2) and rawb=img(:,:,3) respectively, I get this error, ??? Index exceeds matrix dimensions. Anyone knows how I can fix this problem? thanks

Answer 1

[Wayne King]

Are you sure that you have an RGB image that is MxNx3

If you enter

>>whos img

what is returned?

[Natsu]

name img

size 3516x5344

bytes 37579008

class uint16

attributes --

Answer 2

[Image Analyst]

If it's really a raw image from a digital camera, you'll have twice as many green pixels as red or blue pixels (look up Bayer pattern). The format may be very different than a standard RGB image like you'd get from imread() - you'll have to look up the documentation for the function you used to read the raw image. You could have 4 NxM arrays all strung together in one tall 4N by M 2D array, or who knows what.

[Walter Roberson]

Not all digital cameras use Bayer patterns internally, though.

[Image Analyst]

That's true. Just to expand on that....Some have three CCD's, one for each color. Others like the Fuji use a Bayer-like pattern which is hexagonal rather than rectangular. And the Foveon camera uses a single chip RGB co-site sensor where every location has all 3 color pixels at different depths (levels) into the sensor. There is no one single "RAW" format agreed to by everyone. He could have a raw camera image, or he could have a RAW image that someone output from some other program that is simply all the pixels strung together in some fashion, maybe with or without some header info. He needs to find out the actual format of his image, like you also suggested.

[Walter Roberson]

The Foveon was the one I had in mind; I had read about it when it was released, but had lost the name in the intervening time.

Answer 3

[Walter Roberson]

"raw" image formats are manufacturer-specific and sometimes model-specific. Some ad-hoc standards for raw images are evolving, but they are not widespread yet. Do a google search for the raw image format for your camera model.