How to push a first whites (which is not equal to 0) elements from the mask image into the vector array to make line over stored points.

I have scanned my image horizontally and vertically store the first white elements to a vector. And I want to create a vector which will store the first white elements. After storing the points draw line (as shown in the attached figure) over it.

I have already writen a code in C++, but I don't know how to write it in Matlab.

Here I attach a mask image from where i want to extract the while pixels (First white). For clarifications I have pointed out- what does first whites means in the mask image. Any help is much appreciated to achieve my target.

Hope this is clear.

// C++ code with openCV is shown below:
std::vector<int> firstWhite; // std::vector is a sequence container that encapsulates dynamic size arrays
    for(int i=0; i<80; i++)
    {
        for(int j=0; j<img.rows; j++)
        {

            if(mask.at<uchar>(cv::Point(i,j)) != 0)
            {
                firstWhite.push_back(j);
                j = 10000;    //  for example I have 10000 points in the mask image. I need quite fast access to every point.It is much more efficient to allocate too much memory than letting an array grow.
            }
        }

    }

    cv::line(img,cv::Point(0,firstWhite[0]),cv::Point(79,firstWhite[79]),cv::Scalar(0,0,0),2 );
    
// Converting C++ code to matlab. Below is my trial attemp. But unable to achieve the task.

[columns, rows, plane]=size(maskimage);
firstWhite=[]
for i=1: 80
    for j=1:rows
    if not(maskimage(i,j)==0)
        firstWhite(end+1)=j
    end
    end
end

[KALYAN ACHARJYA]

I will try & respond

[kanika bhalla]

@KALYAN ACHARJYA Thank you so much for your help!!

[kanika bhalla]

@Adam Danz@Image Analyst @Walter Roberson, @Guillaume@Matt Jand @KSSV. Could you please help me in this.

Thank you!!

[KSSV]

Attach the original image.

You can get white pixels using:

I = imread('image.png') ; 
I = imcrop(I) ;   % pick only the required part of image 
I1 = rgb2gray(I) ; 
I2 = imbinarize(I1) ; 

[y,x] = find(I2) ; 
imshow(I)
hold on
plot(x,y,'.r')

[kanika bhalla]

Thank you so much for your reply @KSSV.

But I want to extract only first whites from the image (as showed in the image for reference). I had tried this [y,x] = find(I2); earlier. This is returning all whites present in the image.

As my task is to draw line over an image. Therefore, I am focused on finding first whites instead of all whites. Another constraint is that all process should be automatic. All process should be automatic. Even I tried in C++ but I don't know how to achieve that task in matlab. I have also attached the code in this post.

Can you please further guide me on this.

Here I attach an original image for your reference.

Many thanks for your help!!

Answer 1

[Walter Roberson]

iminarize(), bwareafilt()

After that if your binary image is BW, then

first_white_row_idx = sum(cumprod(~BW,1)) + 1

but be careful: rows that have no white at all will show up with output one more than the number of rows

The code above is a vectorized way of find() of the first non-zero pixel for each column. Black is turned to 1's, cumprod that so that the leading 1's on a column are kept but the first 0 stops the run, sum; that gives you the number of 1's before the first 0, which is by construction the number of black pixels before the first white pixel. Add 1 to get the index of the first white pixel.

[kanika bhalla]

Thank you so much @Walter Roberson. I tried this and it has given me some values. But my task is to draw line by extracting the points or positions from the first whites from the image. Can you please guide me how I can use " first_white_row_idx = sum(cumprod(~BW,1)) + 1" to draw line.

As I had already writen code in C++ but I don't know how to achieve this task in Matlab.

Code is shown for your reference:

// C++ code with openCV is shown below:
std::vector<int> firstWhite; // std::vector is a sequence container that encapsulates dynamic size arrays
    for(int i=0; i<80; i++)
    {
        for(int j=0; j<img.rows; j++)
        {
            if(mask.at<uchar>(cv::Point(i,j)) != 0)
            {
                firstWhite.push_back(j);
                j = 10000;    //  for example I have 10000 points in the mask image. I need quite fast access to every point.It is much more efficient to allocate too much memory than letting an array grow.
            }
        }
    }
    cv::line(img,cv::Point(0,firstWhite[0]),cv::Point(79,firstWhite[79]),cv::Scalar(0,0,0),2 );// drawing lines using these coordinates

Thank you so much for your time and help!!

[Walter Roberson]

first_white_row_idx = sum(cumprod(~BW,1)) + 1;
mask = first_white_row_idx > size(BW,1);
first_white_row_idx(mask) = nan;
plot(first_white_row_idx);

[kanika bhalla]

Thank you so much for your help @Walter Roberson

But it didn't work.

Here I attach the input binary and output line drawn image. Could you please help me how to achieve my target of storing first whites and ultimately draw line using those points. Many thanks for your patience and support.

[Walter Roberson]

idx_offset = 1;   %MATLAB indices start from 1

%// C++ code with openCV is shown below:
%std::vector<int> firstWhite; // std::vector is a sequence container that encapsulates dynamic size arrays
firstWhite = zeros(1, 0, 'int32');

%    for(int i=0; i<80; i++)
for i = int32(1) : int32(80)
%    {
%        for(int j=0; j<img.rows; j++)
   for j = int32(0) : int32(img.rows)
%        {
%            if(mask.at<uchar>(cv::Point(i,j)) != 0)
          if mask(i_idx_offset,j+idx_offset)
%            {
%                firstWhite.push_back(j);
             firstWhite(end+1) = j;
%                j = 10000;    //  for example I have 10000 points in the mask image. I need quite fast access to every point.It is much more efficient to allocate too much memory than letting an array grow.
%            }
             break;
          end
%        }

%    }
   end
end

%    cv::line(img,cv::Point(0,firstWhite[0]),cv::Point(79,firstWhite[79]),cv::Scalar(0,0,0),2 );
    
 hold on
 plot([0, 79], firstWhite([0,79]+idx_offset), 'k', 'linewidth', 2);

[kanika bhalla]

Mathworks community is outstanding platform of science and you are leader @Walter Roberson. Thank you so much for dedicating your precious time to help me in my research. I am really thankful to you.

However, I tried to run above code but its giving an error. Here I am attaching slightly modified code. Please help me to achieve my target of drawing line over the first whites.

I have also attached a original, mask, and error image for your reference.

Input_Originalimage = imread('Original_Defectimage'); % Step 1: give original image as an input
Crop_image = imcrop(Input_Originalimage,[360 270 306 185]); % Step 2: Crop the original image

[height, width, plane] = size(Crop_image);

R= Crop_image(:,:,1) % take the red channel of Crop_image
G=Crop_image(:,:,2) % take the green channel of Crop_image
B=Crop_image(:,:,3) % take the blue channel of Crop_image

mask = zeros(height,width); % to create mask image

for i = 1:height
    for j = 1:width
        
      if R(i,j)> 130 & R(i,j) < 150 & G(i,j) > 150 & G(i,j) < 180 & B(i,j) > 110 & B(i,j) < 150
  
          mask(i,j) = 255;    % store 255 pixel values                                                                                                                                                                                          
      end 
      end  
 end

figure; imshow(mask); % this is how I calculated a mask image

% Below is slightly modified code given by you to draw lines over first
% whites
[columns, rows, plane]=size(mask);

idx_offset = 1;   %MATLAB indices start from 1

%std::vector<int> firstWhite; // std::vector is a sequence container that encapsulates dynamic size arrays

firstWhite = zeros(1, 0, 'int32');
%    for(int i=0; i<80; i++)
for i = int32(1) : int32(80)
%    {
%        for(int j=0; j<img.rows; j++)
   for j = int32(0) : int32(rows)
%        {
%            if(mask.at<uchar>(cv::Point(i,j)) != 0)
          if C1(i_idx_offset,j+idx_offset)
%            {
%                firstWhite.push_back(j);
             firstWhite(end+1) = j;
%                j = 10000;    //  for example I have 10000 points in the mask image. I need quite fast access to every point.It is much more efficient to allocate too much memory than letting an array grow.
%            }
             break;
          end
%        }
%    }
   end
end
%    cv::line(img,cv::Point(0,firstWhite[0]),cv::Point(79,firstWhite[79]),cv::Scalar(0,0,0),2 );
    
 hold on
 plot([0, 79], firstWhite([0,79]+idx_offset), 'k', 'linewidth', 2);

[Walter Roberson]

filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/656085/image.jpeg';

Input_Originalimage = imread(filename); % Step 1: give original image as an input
Crop_image = imcrop(Input_Originalimage,[360 270 306 185]); % Step 2: Crop the original image
imshow(Crop_image)

[height, width, plane] = size(Crop_image);

R= Crop_image(:,:,1); % take the red channel of Crop_image
G=Crop_image(:,:,2); % take the green channel of Crop_image
B=Crop_image(:,:,3); % take the blue channel of Crop_image

mask = R > 130 & R < 150 & G > 150 & G < 180 & B > 110 & B < 150;
mask = bwareafilt(mask, [50 inf]);

figure; imshow(mask); % this is how I calculated a mask image

firstWhite = zeros(1,width);
for col = 1 : width
   fw = find(mask(:,col),1);
   if isempty(fw)
     firstWhite(col) = nan;
   else
       firstWhite(col) = fw;
   end
end

 hold on
 plot(firstWhite, 'y', 'linewidth', 2);

However you should be aware that the color range you are identifying is mostly green, only barely touching white (red maximum, green minimum, blue maximum). It includes colors such as Malachite Green, and Olivine. Your code is not designed to detect the yellow-ish stripe. https://www.color-blindness.com/color-name-hue/

[kanika bhalla]

Thank you so much Sir @Walter Roberson. You're really great!!. I can't thank you enough!!!. I was stuck so badly with this one and you solved it within a day!!! I wish I had a brain like you!!!

Can you please further guide me how can I automatically extract the specified coordinates (shown in the attached figure) to draw a line between those positions using -

line([Coordinate1 Coordinate2],[Coordinate'1 Coordinate'2],'color','b') 

I need to draw a line between those points only because my ultimate goal is to do laser cutting by drawing that lines on the original colored defect image.

Thank you so much for your precious time and help.

[Walter Roberson]

filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/656085/image.jpeg';

Input_Originalimage = imread(filename); % Step 1: give original image as an input
Crop_image = imcrop(Input_Originalimage,[360 270 306 185]); % Step 2: Crop the original image
imshow(Crop_image)

[height, width, plane] = size(Crop_image);

R= Crop_image(:,:,1); % take the red channel of Crop_image
G=Crop_image(:,:,2); % take the green channel of Crop_image
B=Crop_image(:,:,3); % take the blue channel of Crop_image

mask = R > 130 & R < 150 & G > 150 & G < 180 & B > 110 & B < 150;
mask = bwareafilt(mask, [50 inf]);

figure; imshow(mask); % this is how I calculated a mask image

firstWhite = zeros(1,width);
for col = 1 : width
   fw = find(mask(:,col),1);
   if isempty(fw)
     firstWhite(col) = nan;
   else
       firstWhite(col) = fw;
   end
end

 hold on
 plot(firstWhite, 'y', 'linewidth', 2);

 
 figure;
 dfw = diff(firstWhite);
 plot(dfw)
 title('difference in firstWhite')

 left_edge_x = find(dfw > 3, 1);
 left_edge_y = firstWhite(left_edge_x);
 right_edge_x = find(dfw < -3, 1, 'last')+1;
 right_edge_y = firstWhite(right_edge_x);
 figure
 imshow(mask)
 hold on
 plot(firstWhite, 'y', 'linewidth', 2);
 plot([left_edge_x, right_edge_x], [left_edge_y, right_edge_y], 'color', [1 .5 0], 'linewidth', 2)

[kanika bhalla]

I am indebted to your kindness and knowledge. I don't know of I'll ever be able to repay you properly. I am forever grateful to you Sir @Walter Roberson.

Next task is to find the first whites from the bottom to draw a line. I am trying best to implement the same according to your provided knowledge and solutions.

Thank you so much for your valuable time and support.