Re: Blood vessel analysis using 16-bit image

Posted by Kenneth Sloan-2 on Jul 28, 2009; 7:24pm
URL: http://imagej.273.s1.nabble.com/Blood-vessel-analysis-using-16-bit-image-tp3691611p3691619.html

On Jul 28, 2009, at 1:59 PM, Dianne Patterson wrote:

> It matters how you convert to 8 bit....you can lose all values above  
> 256, if
> you don't take pains to preserve them.
> I do  the following
> # copy the 16 bit image to a new8
> # Find the maximum intensity value in the image and put it in a  
> variable
> # Divide max by 255
> # Use divisor to change the range of values in new8, then save it as  
> 8-bit
> char
>
> When you are done, look at the old image and the new one...they  
> should look
> the same to you...and white values should be near 255.
>
> -Dianne

That's just the first step.  You might also consider:

a) finding the Min value and mapping [Min,Max] -> [0,255]  (note that  
this might INCREASE the
    range of values)
b) computing further statistics, and performaing a more
    general transformation of the values

It all depends on whether or not the values themselves have any  
intrinsic meaning (other than "darker" and "lighter").  But, that ship  
already sailed when you re-scaled to fit in [0,255].


The point is: as soon as you decide that you have to modify the actual  
pixel values, you must have some information about what the pixels  
*mean* and how they will be processed in the next step.


--
Kenneth Sloan
[hidden email]