Extract XY data from the image based on the color filter

Hi,

How could I extract XY data from the image by selecting points of the same color? For example, in the attached figure, gain is plotted versus frequency for three different temperatures. +85C is red color, +25C is green color and -40C is blue color. I would like to be able to extract XY data but instead of picking up manually point by point on the curve, I would like to pick one point on the red curve which would then select all points with the same color and export them as the XY data scaled based on the some already known XY points presented on the graph.

Thank you,

S.R.

[Ameer Hamza]

Do you just have an image, or do you also have a .fig file?

[S.R.]

Hi Ameer,

It is just an Image. I know there are some digitizer codes but I am looking for simple solution becasue new datasheets always come in colors. Therefore, I would like to know if there is a simple solution that would allow for extracting XY data by selecting one point for example on red line (indicating that data in red should be extracted) and extracting XY data automatically versus clicking point after point.

Thank you,

[Ameer Hamza]

Try this package: https://www.mathworks.com/matlabcentral/fileexchange/7173-grabit

[S.R.]

Hi Ameer,

I tried it, but "Grabit" does not have a color filter or in other words there is no ability to detect color in the image and extract data based on the color.

Thank you,

[Ameer Hamza]

I think there is no straightforward way to extract the data. You may need to write the code according to your requirement using the image processing tools.

[S.R.]

Hi Ameer,

I ma looking into this function "impixelinfoval" that returns pixel location and color when I hover over the attached figure. If I could return this information back, I assume that there is a function for locating or scanning entire picture and returning pixel locations with the same color?

h = imshow('new_LTC5596.png')

hText= impixelinfoval(gcf,h);

set(hText,'FontWeight','bold')

set(hText,'FontSize',10)

Thank you,

Answer 1

[Image Analyst]

Try imsplit

[r,g,b] = imsplit(rgbImage);
% Then scan each column with find() until you find the first row where that color appears.
redImage = r == 255 & g == 0 & b == 0;
greenImage = r == 0 & g == 255 & b == 0;
blueImage = r == 0 & g == 0 & b == 255;
[rows, columns] = size(redImage)
ry = zeros(1, columns)
gy = zeros(1, columns)
by = zeros(1, columns)
for col = 1 : columns
    % First the red
    thisCol = redImage(:, col);
    topRow = find(thisCol);
    if ~isempty(topRow)
        ry = topRow;
    end
    % Next the green
    thisCol = greenImage(:, col);
    topRow = find(thisCol);
    if ~isempty(topRow)
        gy = topRow;
    end
    % Next the blue
    thisCol = blueImage(:, col);
    topRow = find(thisCol);
    if ~isempty(topRow)
        by = topRow;
    end
end

Of course the values will be in units and coordinates of image pixels, so you'll have to calibrate the distances.

[S.R.]

Hi Image Analyst,

I am using "impixelinfoval" to get pixels and color which would probably be the first step. How can I return this information back to the code ?

Example figure is attached in previous response to Ameer above.

h = imshow('new_LTC5596.png')

hText= impixelinfoval(gcf,h);

set(hText,'FontWeight','bold')

set(hText,'FontSize',10)

Thank you,

[Image Analyst]

Are you sure you can't get the figure, or the data used to create the figure. If you have just the PNG image, it's going to be hard to pull out each curve independently since they overlap. You're probably best off using imfreehand to hand trace the curve. I'm attaching a demo.

[S.R.]

Hi Image Analyst,

Absolutely sure. When doing RF design we normally work from the datasheet. Newer datasheets come in colors and the old ones come in black and white. I use WebPlotDigitizer but would like to implement this type of code in Matlab due to the other scripts I am using. If the plots overlap I see that as a problem but in that case I would use manual type of digitizer where I select points. In the case where we can extract datapoints, this more automated way would be of tramnedous help becasue it would speed up process.

I am getting pixel information and RGB color now with the code below but I am not certain how to pass variables to your code to be able to find matching RGB colors and re-plot figure?

Thank you for sending freehand drawing demo. I will check it out.

rgbImage = imread('new_LTC5596.png');

h=imshow(rgbImage);

msgbox({'You can zoom and pan before selecting each point';'press "space" when you are done to start the selection';'If you made a mistake, you can delete latest points pressing "backspace"';'Once you are finished, press "enter" to exit'})

[x,y] = ginput_zoom();

% Put a cross where they clicked.

hold on;

plot(x, y, 'w+', 'MarkerSize', 200);

% Get the location they click on.

% Extract the RGB color from that location.

[row,col,rgb]=impixel(rgbImage,x,y)

function [x,y] = ginput_zoom()

x=[];y=[];

h=[];

w=1;

while w~=0;

w = waitforbuttonpress;

while w==0

w = waitforbuttonpress;

end

cfg = gcf();

ch = double(get(cfg, 'CurrentCharacter'));

if ch == 13 % ENTER button

break;

end

if ch == 8 % Backspace button

if isempty(x) == 0

x = x(1:end-1);

y = y(1:end-1);

delete(h(end));

h = h(1:end-1);

continue;

end

end

[a,b]=ginput(1);

x = [x;a];

y = [y;b];

hold on; h = [h;plot(x,y,'g+')]; hold off;

end

Thank you,