How do I get the values of the surrounding pixels when I choose a particular pixel in an image?

I'm trying to estimate the possible points of intersection between two lines (arrays) in an image. Is there any function to return the values of the surrounding pixels for a particular value of the array?

Answer 1

[Image Analyst]

Assuming it's a gray scale image, and that the row and column is not on the edge of the image:

surroundingWindow = grayImage(row-1:row+1, col-1:col+1);

If it's color:

surroundingWindow = rgbImage(row-1:row+1, col-1:col+1, :);

[The Imagician]

I don't understand how this works, can you please explain? Let's say I use impixel to get the coordinates of the pixel. How does this work?

[The Imagician]

The green arrow is part of the image. Whereas the yellow line is a plot of points. So when I try to find their intersection, I don't get the any points which I can visually see. So i'm trying to estimate the intersection by finding the neighbors of each point or coordinate of the yellow line. Is there any other approach you would suggest?

[Image Analyst]

When you said "for a particular value of the array" I assume you had the row and column and when you said "values of the surrounding pixels" I assume you meant the pixel values in a 3-by-3 square around that particular row and column.

For the case you gave now (which you should have posted in the beginning to save time), you must have the coordinates of the yellow and green lines, or else you could not have plotted them. If you have the Statistics and Machine Learning Toolbox, you can use pdist2() to get the distance of every yellow point to every green point. Then you can threshold that to find out which are closer than a certain threshold. You need to do this because there's a good chance that there are no points which have the exact values out to dozens of decimal points. So using pdist2() you can find out which yellow and green points (the indexes) are very close to each other.

Give it a try yourself first. If you can't do it, post some data and I'll see what I can do.

[The Imagician]

I have 38 yellow points and 752 green points. I have calculated the pdist2() between every yellow point with every green point and got a 38x752 matrix. I used the below code to NaN values greater than 50 from this matrix. Hope my approach to this is in parallel with your suggestion.

Pdist = pdist2(newdistortion,greenArrow,'euclidean');
indices = find(abs(Pdist)>50);
Pdist(indices) = NaN;

[Image Analyst]

You'd need to do this:

[rows, columns] = find(abs(Pdist) > 50);

[The Imagician]

Thank you! The logic is great and helped me a lot :-)