ImageJ

how to count the number of peaks in a dynamic profile

_‹ Previous Topic Next Topic _›

Classic

List

Threaded

6 messages Options

paul_rostand

how to count the number of peaks in a dynamic profile

Hello imagej users,

I am trying to count and locate the peaks in a picture.for this purpose I use the following macro wich is an adaptation of the FindMaxPixelLocation macro. for peak identification I just compare the values of consecutives points

(curve[i+1]) && (curve[i+1]>curve[i+2])

curve[i] being the pixel value at location i of the loop. but as result I get 39 peaks instead of 4 (see the attached profile)

I would like to know what I am doing wrong and if anyone has another easier way to solve this problem

thanks in advance

regards

Paul Rostand

   showStatus("Finging pixel with largest value...");
   max = 0;
    peak=0;
   width = getWidth();
height = getHeight();
   for (y=0; y<height; y++) {
      if (y%20==0) showProgress(y, height);
      for (x=0; x<50; x++) {
          value = getPixel(x,y);
          if (value>max) {
              max = value;
              xmax = x;
              ymax = y;
          }
      }
}
print("A pixel with the max value ("+max+") is located at "+xmax+","+ymax);

curve=newArray(width);
            threshold = 0.9*max;           // define the threshold used to count the peaks

         for (y=0; y<height; y++) {
if (y%20==0) showProgress(y, height);
            for (x=0; x<50; x++) {
                 value = getPixel(x,y);
   if (value>threshold) {
   for (i=0;i<48;i++) {
   curve[i]=value;
             if((curve[i]<curve[i+1]) && (curve[i+1]>curve[i+2])){
                                                  peak=peak+1;
   }
}
          }
      }
}

print("the number of peaks is "+peak+");

FindMaxPixelLocation.txt (1K) Download Attachment

Profile of in_0030.jpg (18K) Download Attachment

lechristophe

Re: how to count the number of peaks in a dynamic profile

Dear Paul,

There is a Process>Find Maxima command in ImageJ that works quite well to
detect peaks in images. Did you try to use it ?

Christophe

On Thu, Apr 28, 2011 at 11:19, paul_rostand <[hidden email]> wrote:

> Hello imagej users,
>
> I am trying to count and locate the peaks in a picture.for this purpose I
> use
> the following macro wich is an adaptation of the FindMaxPixelLocation
> macro. for
> peak identification I just compare the values of consecutives points
> (curve[i+1]) && (curve[i+1]>curve[i+2])
>
> curve[i] being the pixel value at location i of the loop. but as result I
> get 39
> peaks instead of 4 (see the attached profile)
>
> I would like to know what I am doing wrong and if anyone has another easier
> way
> to solve this problem
> thanks in advance
> regards
>
> Paul Rostand
>
> showStatus("Finging pixel with largest value...");
> max = 0;
> peak=0;
> width = getWidth();
> height = getHeight();
> for (y=0; y<height; y++) {
> if (y%20==0) showProgress(y, height);
> for (x=0; x<50; x++) {
> value = getPixel(x,y);
> if (value>max) {
> max = value;
> xmax = x;
> ymax = y;
> }
> }
> }
> print("A pixel with the max value ("+max+") is located at "+xmax+","+ymax);
>
> curve=newArray(width);
> threshold = 0.9*max; // define the threshold used to
> count
> the peaks
>
> for (y=0; y<height; y++) {
> if (y%20==0) showProgress(y, height);
> for (x=0; x<50; x++) {
> value = getPixel(x,y);
> if (value>threshold) {
> for (i=0;i<48;i++) {
> curve[i]=value;
> if((curve[i]<curve[i+1]) && (curve[i+1]>curve[i+2])){
> peak=peak+1;
> }
> }
> }
> }
> }
> print("the number of peaks is "+peak+");
>
> --
> View this message in context:
> http://imagej.588099.n2.nabble.com/how-to-count-the-number-of-peaks-in-a-dynamic-profile-tp6312558p6312558.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>

suendermann

Re: how to count the number of peaks in a dynamic profile

In reply to this post by paul_rostand

Hello Paul,

maybe I'm wrong, but I think the problem is that your detection is too
sensitive. I would take the values given by the plot profile and filter
them with a 1D median or mean filter to get rid of the "noisy peaks".
On the other hand you could try to enlarge your comparison of the left
and the right neighbour to the 2 left and the 2 right neighbours. There
you only get the real big maxima.

Greeting
Fred

On 28.04.2011 11:19, paul_rostand wrote:

> Hello imagej users,
>
> I am trying to count and locate the peaks in a picture.for this purpose I use
> the following macro wich is an adaptation of the FindMaxPixelLocation macro. for
> peak identification I just compare the values of consecutives points
> (curve[i+1])&& (curve[i+1]>curve[i+2])
>
> curve[i] being the pixel value at location i of the loop. but as result I get 39
> peaks instead of 4 (see the attached profile)
>
> I would like to know what I am doing wrong and if anyone has another easier way
> to solve this problem
> thanks in advance
> regards
>
> Paul Rostand
>
> showStatus("Finging pixel with largest value...");
> max = 0;
> peak=0;
> width = getWidth();
> height = getHeight();
> for (y=0; y<height; y++) {
> if (y%20==0) showProgress(y, height);
> for (x=0; x<50; x++) {
> value = getPixel(x,y);
> if (value>max) {
> max = value;
> xmax = x;
> ymax = y;
> }
> }
> }
> print("A pixel with the max value ("+max+") is located at "+xmax+","+ymax);
>
> curve=newArray(width);
> threshold = 0.9*max; // define the threshold used to count
> the peaks
>
> for (y=0; y<height; y++) {
> if (y%20==0) showProgress(y, height);
> for (x=0; x<50; x++) {
> value = getPixel(x,y);
> if (value>threshold) {
> for (i=0;i<48;i++) {
> curve[i]=value;
> if((curve[i]<curve[i+1])&& (curve[i+1]>curve[i+2])){
> peak=peak+1;
> }
> }
> }
> }
> }
> print("the number of peaks is "+peak+");
>
> --
> View this message in context: http://imagej.588099.n2.nabble.com/how-to-count-the-number-of-peaks-in-a-dynamic-profile-tp6312558p6312558.html
> Sent from the ImageJ mailing list archive at Nabble.com.

--
Frederik Sündermann, Dipl.-Biol.
Department of Neurobiology
University of Osnabrück
Barbarastraße 11
49076 Osnabrück
Tel: +49 541 969-2846

paul_rostand

Re : how to count the number of peaks in a dynamic profile

In reply to this post by lechristophe

thank you Fred and Christophe,

<<<fred ; increace the step of iteration just increase the amount of peaks

<<<Christophe : the problem with Process>Find Maxima ist that i have a file with more than 1000pictures with and without holes in the middle and all with different noise tolerance. and I want to use the fact that the hole is always in the middle of 2 maximas as criteria to sort my pictures before stacking.

regards

De : lechristophe [via ImageJ] <[hidden email]>
À : paul_rostand <[hidden email]>
Envoyé le : Jeu 28 avril 2011, 12h 12min 01s
Objet : Re: how to count the number of peaks in a dynamic profile

Dear Paul,

There is a Process>Find Maxima command in ImageJ that works quite well to
detect peaks in images. Did you try to use it ?

Christophe

On Thu, Apr 28, 2011 at 11:19, paul_rostand <[hidden email]> wrote:

If you reply to this email, your message will be added to the discussion below:

http://imagej.588099.n2.nabble.com/how-to-count-the-number-of-peaks-in-a-dynamic-profile-tp6312558p6312709.html

To start a new topic under ImageJ, email [hidden email]
To unsubscribe from ImageJ, click here.

Michael Schmid

Re: how to count the number of peaks in a dynamic profile

In reply to this post by paul_rostand

Hi Paul,

for finding whether a point is a maximum it is not enough to check
whether a profile along the x direction has a maximum. It must be
also a maximum in the y direction (and in the diagonal directions,
when considering 8-connected pixels).

Also, you may have a maximum with several pixels of the same value,
which would be lost in your approach.

So, I'd recommend ImageStatistics for finding the overall maximum
(this is much faster than the macro approach) and Process>Find Maxima
for finding local maxima. "Find Maxima" also has an option to limit
maximum finding to values above the lower threshold of a thresholded
image - just threshold your image between 90% of the overall maximum
and the full maximum.

Michael
________________________________________________________________

On 28 Apr 2011, at 11:19, paul_rostand wrote:

> Hello imagej users,
>
> I am trying to count and locate the peaks in a picture.for this
> purpose I use
> the following macro wich is an adaptation of the
> FindMaxPixelLocation macro. for
> peak identification I just compare the values of consecutives points
> (curve[i+1]) && (curve[i+1]>curve[i+2])
>
> curve[i] being the pixel value at location i of the loop. but as
> result I get 39
> peaks instead of 4 (see the attached profile)
>
> I would like to know what I am doing wrong and if anyone has
> another easier way
> to solve this problem
> thanks in advance
> regards
>
> Paul Rostand
>
> showStatus("Finging pixel with largest value...");
> max = 0;
> peak=0;
> width = getWidth();
> height = getHeight();
> for (y=0; y<height; y++) {
> if (y%20==0) showProgress(y, height);
> for (x=0; x<50; x++) {
> value = getPixel(x,y);
> if (value>max) {
> max = value;
> xmax = x;
> ymax = y;
> }
> }
> }
> print("A pixel with the max value ("+max+") is located at "+xmax
> +","+ymax);
>
> curve=newArray(width);
> threshold = 0.9*max; // define the threshold
> used to count
> the peaks
>
> for (y=0; y<height; y++) {
> if (y%20==0) showProgress(y, height);
> for (x=0; x<50; x++) {
> value = getPixel(x,y);
> if (value>threshold) {
> for (i=0;i<48;i++) {
> curve[i]=value;
> if((curve[i]<curve[i+1]) && (curve[i+1]>curve[i+2])){
> peak=peak+1;
> }
> }
> }
> }
> }
> print("the number of peaks is "+peak+");
>
> --
> View this message in context: http://imagej.588099.n2.nabble.com/
> how-to-count-the-number-of-peaks-in-a-dynamic-profile-
> tp6312558p6312558.html
> Sent from the ImageJ mailing list archive at Nabble.com.

Gabriel Landini

Re: how to count the number of peaks in a dynamic profile

On Thursday 28 Apr 2011, Michael Schmid wrote:
> for finding whether a point is a maximum it is not enough to check
> whether a profile along the x direction has a maximum. It must be
> also a maximum in the y direction (and in the diagonal directions,
> when considering 8-connected pixels).
> Also, you may have a maximum with several pixels of the same value,
> which would be lost in your approach.

In addition to Michael's comments, let me say that this sort of problem has
been resolved using mathematical morphology (in particular greyscale
reconstruction) long time ago and there are formal ways of doing it.

What I guess you are looking for is called "regional maxima" and a related
procedure "h-domes". Reading about this will save you lots of time in getting
the algorithms right. The papers by Luc Vincent (many of those are online)
will cover this in detail and with examples.
Pierre Soille's book (Morphological image analysis) is also very good on
this.
Several of these procedures have been implemented in the Morphology plugin
collection available in my page.

Cheers

Gabriel