IDF with IMAQ?

[Atomic]

Hi, I am a graduate student using the department's copy (site liscensed I am told) of IMAQ vision for Labview (we are using IMAQ 7 with Labview 6).

My labview skills are primitive but I managed to make some image processing software for my research project (blob analysis sort of thing).

In my project I collect images from a microscope camera wherein each image is taken when the objective of the scope is at a slightly different distance from the specimine, thus working like a confocal scope, except we are looking at darkfeild plan view with reflected light. The 'stacks' of images number between 60 and 120, of them all but 1 to 5 are through away images (nothing to see ... reason being that the objects we are looking at change size over the course of the experiment and only by rastoring the focal plane so dramatically (@~2 microns/step) do we always maintain at least some of the object in focus overnight). Of the 1 to 5 images that remain that show anything of interest, I usually just choose 1 with the best focus of the object; however, there are instances when the objects features are captured in focus over 2 to several images (ie. each image has a bit in focus). I have played with software from Alicona that is called IDF (infinite depth of field) wherein you take a stack of such images each with a bit of an object in focus and the Alicona software makes a single image that is completely in focus. The Alicona software doesn't work very well, and I wish I could tweak it like I can my blob analysis software composed in labview using IMAQ vision. Therefore, I would like to make such an app using labview and IMAQ vision, one that will both automatically scrub the images that are of no use (nothing in focus) and then combine the remaining images into a single image that shows all of the objects in the field in focus.

Please help.

[Jeffrey_P]

Good afternoon Atomic,

There aren't any built in functionalities to NI Vision that can do IDF.  I'm not that familiar with the algorithms; however, I'm sure that the actual algorithm used could be found.  I would imagine that it would entail creating individual masks off of each image from that which is in focus, and then combining each individual mask into a comprehensive image.  The most challenging part of this would be the determining of what is "in focus", and then creating the respective mask from that "in focus" image.  There aren't any built-in functionalities or examples to get you started, though!

If you have questions along the way about how to do certain features, please feel free to post back with more direct questions!

Respectfully,

[Atomic]

I have read about this masking business, I think I can figure that part out. The real trick is going to be the determination of 'in focus'.

To summarize for those who have not ready the original message: - I need to have software determine which images among many have a part that is 'in focus', and furthermore, what part of the image that is 'in focus'.

Any help is appreciated.

[chilly_charly]

When an object is in focus, its edges are sharp : means that the local contrast (gradient) is maximum. So may be you should extract the pixels with the higest gradient.

Could you keep us informed about your progress ?

Message Edité par chilly charly le 07-10-2006 06:27 AM

[chilly_charly]

Seems that the Laplacian is the best tool to define the focus quality. Use the attached image a see the the result of applying a 3x3 Laplacian filter

Message Edité par chilly charly le 07-10-2006 06:41 AM