Stacking tif images of wafers

Question

Hi all,

Does anyone have experience in stacking tif images and using the images analyzer to convert it into a JMP file? My issues are:

* I don’t know how to stack images, so I manually join the tables after converting 
* The images are 1.5 mb giving 2.4 million rows of data per picture making it cumbersome to work with

The customer wishes to stack/analyze as many as pictures as possible to look for error patterns in JMP.

Background:

The customer is in the semiconductor industry and is producing wafers. They wish to buy JMP but worries that their data isn’t suitable for analysis. They wish to discover error patterns in their wafers and I showed them the hierarchical clustering function on the “Wafer stacked” data sheet that comes with JMP (thank you Mike). This is exactly what they wish to achieve with their own wafers.

They scan the surface using scanning acoustic microscope and get large images in the tif format.

Thanks!

MikeD_Anderson · Accepted Answer

Hi, Christian!

The best way to create the stacks is through ImageJ.  There's a great bundled version of ImageJ called Fiji (Fiji Is Just ImageJ).  You can get it here: Fiji Is Just ImageJ

Then you just want to create an image stack.  A tutorial for that can be found here (that first page is a little oddly formatted, so I put in a couple of videos, too).  Note that the first video gives a faster way to import the images than the second.  The first video also has the advantage of showing a method that allows you to bin the images down during import (you want to test it first).  That will significantly reduce the data set size without much loss in data integrity.  The second video shows a long way around but shows how to save out the tiff image stack.

Creating Tiff Stacks

https://www.youtube.com/watch?v=dL1piUUNjsw

https://www.youtube.com/watch?v=Sv00y8E6vdA

Hope that helps!

Best,

M

MikeD_Anderson · Accepted Answer

Big Images.  If you start off with a 4MP image that’s going to be ~8 million rows for two images.  If you down-sample that by 50% to 2MP you’re still going to be looking at 4 Million rows.  You can down-sample quite far if you use the longer stacking method (one the links I sent earlier) and use a sum or max statistic to aggregate the pixels.

Best,

M