Legacy ImageMagick Discussions Archive

Hi,

I have 3 questions.

1.
I have 3 separate tif images (16-bit pixel depth each) representing data acquired with a scanner.
Every single file is a grayscale image corresponding to only one of three R, G and B channels: RED.tif, GRN.tif, BLU.tif (in fact there is a 4th image acquired within near infra-red; NIR.tif).
I'd like to "reassemble" or "merge" them in a single RGB tif image with no data loss, i.e. = 3*16=48-bit total.

By doing : The resulting RGB merged.tif image does have some "flat regions" (i.e. single color for area where the color details were not high, like the image was segmented) where I feel like there was some strong compression like we can somehow see on JPEG images.
Adding the -compress none flag doesn't affect the result at all.
How could I avoid this quality loss?



2nd question:
The 3 original images are not perfectly aligned. There are some pixels shift.
What I would achieved is that the first image taken is the reference (top-left corner = [0,0]) and the two others may have some shift.

For example let's say we want a shift along x (horizontal axis) and y (vertical axis):
GRN.tif: +5px along x and +3px along y, thus, top-left corner would be [5,3] in the reference axis of the first image (RED.tif)
BLU.tif: -7px along x and -4px along y, thus, top-left corner would be [-7,-4] in the reference axis of the first image (RED.tif). Here I want to lose a 7 pixels band on the left of the BLU.tif and a 4 pixels band on top.

Here are the images dimensions in pixels (width x height):
NIR.tif: 693 x 6820
RED.tif: 693 x 6825
GRN.tif: 693 x 6825
BLU.tif: 693 x 6823

They are by chance here the same in some cases, but It's not always the case.
No matter what append and the bottom and the right side (but I would probably have to manage that after having solved these first issues).


3rd question:
Remember I also have a 4th image taken with an infra-red sensor:
NIR.tif

Is there something special to know to achieve some other combinations of channels?
For example, a NRG false color image, i.e. N stands for near infrared, R for red and G for green original channels.
So, what I call a NRG image is an image where these 3 original channels are set, or "projected" into the RGB "color representation" needed for display on a screen monitor.
Thus, N becomes R in RGB "color representation", R->G and G->B.
All other thinkable combinations are also possible, like GNB or whatever, but the first image use would always be the reference for geometric alignment as presented before.


Thanks a lot for your precious advices.
Best Regards.

picture sample normally available here
I also dropped an image processed with Photoshop so that you can see the difference in green regions of what I called "flat region". Zoom 100% required to clearly see that.

Your link doesn't work for me.

You can crop before combining, eg: If you list them in a diffeent order before "-combine", this gives you false channels.

Thanks, I'll try it tomorrow, on bash/unix it doesn't work (syntax error). I hope it will works fine on the Windows machine.

Question: what if the offset are negative?
Because it may be the case sometimes...

In bash you need to escape the parentheses \( and \).

Adjust the offsets so that none are negative (unless you want strips down sides where not all channels are used).