Hi Duleep,

first you need a selection of the low-variance areas. Duplicate the

image, use Process>Filters>Variance, smooth it a bit and threshold to

get the low-variance regions. Use Edit>Selection>Create selection and

add it to the ROI manager.

Then you can take the original image, apply the selection from the ROI

Manager (click on it), and use the "Fit Polynomial" plugin to subtract a

polynomial, where the fit is done only in the selected area:

http://imagejdocu.tudor.lu/doku.php?id=plugin:filter:fit_polynomial:startAll these operation s should be done one floating-point (32-bit) images;

if you have different bit depth, use Image>Type to convert to 32 bits

before starting.

Note that this approach works for a very limited class of images only;

there must be a low-variance phase dominating the image. If you have

e.g. a bright low-variance area at the left and a low-variance area from

a different phase (dark) at the right, it will fail.

I tried with the image in Fig. 1 of the article you have mentioned and

the background crated by this algorithm was not good enough for for

segmentation based on the pixel value. I had to manually select the

areas of the majority phase before running "fit polynomial".

(You can apply 'fit polynomial' several times without losing anything,

so the algorithm might be a first step, then it becomes easier to see

which areas belong to the same physe. Eventually, you will reach a state

where you can threshold the majority phase, create a selection from it,

correct it manually, and run "fit polynomial" again).

One more thing you could try is not subtracting the background but

dividing by it (before duplicate the image before "fit polynomial" and

use it's "Output Fit" option, then use Process>Image Calculator). This

might produce better results than subtracting a background.

Michael

________________________________________________________________

On 14/05/2018 08:53, Duleep Samuel wrote:

> Respected ImageJ users, i read in the paper

> Color (and 3D) for Scanning Electron Microscopy,

> doi: 10.1017/S1551929518000482

>

> how they had done

>

> applying mathematical correction of the gray levels. This

> type of image processing consists of

> subtracting the 2nd degree polynomial

> that best fits (least square method) the

> homogeneous (low variance) areas of

> the gray-level image. In the corrected

> image, each color represents a different

> mineral phase that can yield a quantitative

> volume fraction.

>

> How can I do this in ImageJ, I would like to do this work in black and

> white Scanning Electron Microscope Images. Thanks, with regards,

> Dr.D.K.Samuel, IIHR, Bangalore, India

>

> --

> ImageJ mailing list:

http://imagej.nih.gov/ij/list.html>

--

ImageJ mailing list:

http://imagej.nih.gov/ij/list.html