Plotting multiple matrices gnuplot - animation

I'm fairly new to gnuplot and I'm trying to see how my array varies with iterations of my program to debug, I would ideally like to be able to make an animation of the data.
Here's what my data looks like:
'q=0'
1 0 0
0 1 1
1 1 0
'q=1'
1 0 0
0 1 1
0 1 0
and so on.
I tried using: plot "matrix.dat" index 0 matrix with image, just to plot the first matrix but I got "Warning matrix contains missing or undefined values, Matrix does not represent a grid".

I think the problem may be the comment lines. When I used this version of your data file
1 0 0
0 1 1
1 1 0
1 0 0
0 1 1
0 1 0
plotting with index works. To make an animation, you could create a series of .png files and stitch them together with another application. An example of the gnuplot code to make the .pngs would be:
set terminal png
do for [i=0:100] {
set output sprintf('matrix%03.0f.png',i)
plot 'data.dat' index i matrix with image
}

Related

Set-up animation time in paraview from one column

I have a .txt file containing columns of data. At each row, i have 5 columns, containing object id, time, x position, y position, z position.
Example:
ID Time X Y Z
0 0 0 0 0
1 0 1 0 0
3 0 0 1 0
0 1 0 0 0
1 1 0 1 0
I don't know how to set up Paraview to read the column Time as the Time in the animation.
I know it's possible with multiple files (one for each time step), but is it possible with one single file containing all the time steps as shown above ?
Thanks for advice
Not supported in ParaView (last release 5.6.0) by default, but it should be quite easy to do it with a live programmable source.

Connected component labeling in matrix

I'm trying to do the following
Given the following matrix (where 1's are empty cells and 0's are obstacles):
0 0 1 1
1 0 0 0
1 0 1 1
1 1 0 0
I want it to become like this:
0 0 1 1
2 0 0 0
2 0 2 2
2 2 0 0
What I need to do is to label all connected components (free spaces).
What I already tried to do is to write a function called isConnected() which takes indexes of two cells and checks if there is a connected path between them. And by repeating this function n^2 times on every empty cell on the matrix I can label all connected spaces. But as this algorithm has a bad time complexity (n^2*n^2*O(isConnected())) I prefer to use something else.
I hope these pictures will explain better what I'm trying to accomplish:

do image segmentation given the boundary of regions in the image

I get an image and also the boundary of regions in the image. For example, I have a mask with logical type, the value of boundary is 1, while for other pixels, the value is 0. I want to label the regions segmented by the boundaries, while I am not sure how to segment and label the region based on the continuous boundary.
The boundary looks like this:
0 0 0 1 0 0 0 1 0 0
0 0 1 0 0 0 0 1 0 0
1 1 0 1 0 0 0 1 0 0
0 0 0 0 1 0 1 0 0 0
With the above diagram, there would be four regions that would be identified.
The function bwlabel from the image processing toolbox is the ideal function you should use to label each continuous region of non-zero pixels in a binary mask. However, you want to perform this on the zero pixels that are delineated by the "boundary" pixels that are set to 1. Therefore, simply use the inverse of the binary mask so that you are operating on the zero pixels instead of the non-zero pixels. Also from your definition, regions are separated using a 4 pixel connectivity. bwlabel by default uses 8 pixel connectivity when searching for continuous regions, which means that it looks in the N, NE, E, SE, S, SW, W and NW directions. You'll want to manually specify 4 pixel connectivity, which looks in only the directions of N, E, S and W.
Supposing your mask was stored in the variable L, simply do:
labels = bwlabel(~L, 4);
The output labels would be a map that tells you the membership of each pixel. Regions of the same membership tells you that those pixels belong to the same group.
Using your example, we get:
>> L = [0 0 0 1 0 0 0 1 0 0
0 0 1 0 0 0 0 1 0 0
1 1 0 1 0 0 0 1 0 0
0 0 0 0 1 0 1 0 0 0];
>> labels = bwlabel(~L, 4)
labels =
1 1 1 0 3 3 3 0 4 4
1 1 0 3 3 3 3 0 4 4
0 0 2 0 3 3 3 0 4 4
2 2 2 2 0 3 0 4 4 4
Each island of zeroes has a unique ID where pixels that belong to same ID belong to the same island or region. If you don't want to use bwlabel and do this from first principles, you can refer to my previous post using Depth First Search to find regions of connected components: How to find all connected components in a binary image in Matlab?. Be advised that this is not efficient code and so you should only use it for educational and research purposes. Using bwlabel is recommended as it is a fast function and well tested. You'll also have to modify the code so that it doesn't search in an 8 pixel connectivity and it should look at only a 4 pixel connectivity. Make sure you also inverse the input before using the code.

Explanation of Matlab's bwlabel,regionprops & centroid functions

I have spent all day reading up on the above MATLAB functions. I can't seem to find any good explanations online, even on the MathWorks website!
I would be very grateful if anyone could explain bwlabel, regionprops and centroid. How do they work if applied to a grayscale image?
Specifically, they are being used in this code below. How do the above functions apply to the code below?
fun=#minutie; L = nlfilter(K,[3 3],fun);
%% Termination LTerm=(L==1);
figure; imshow(LTerm)
LTermLab=bwlabel(LTerm);
propTerm=regionprops(LTermLab,'Centroid');
CentroidTerm=round(cat(1,LTerm(:).Centroid));
figure; imshow(~K)
set(gcf,'position',[1 1 600 600]); hold on
plot(CentroidTerm(:,1),CentroidTerm(:,2),'ro')
That's quite a mouthful to explain!... nevertheless, I'd love to explain it to you. However, I'm a bit surprised that you couldn't understand the documentation from MathWorks. It's actually quite good at explaining a lot (if not all...) of their functions.
BTW, bwlabel and regionprops are not defined for grayscale images. You can only apply these to binary images.
Update: bwlabel still has the restriction of accepting a binary image but regionprops no longer has this restriction. It can also take in a label matrix that is usually output from bwlabel as well as binary images.
Assuming binary images is what you want, my explanations for each function is as follows.
bwlabel
bwlabel takes in a binary image. This binary image should contain a bunch of objects that are separated from each other. Pixels that belong to an object are denoted with 1 / true while those pixels that are the background are 0 / false. For example, suppose we have a binary image that looks like this:
0 0 0 0 0 1 1 1 0 0
0 1 0 1 0 0 1 1 0 0
0 1 1 1 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 1 1
0 0 1 1 1 1 0 0 1 1
You can see in this image that there are four objects in this image. The definition of an object are those pixels that are 1 that are connected in a chain by looking at local neighbourhoods. We usually look at 8-pixel neighbourhoods where you look at the North, Northeast, East, Southeast, South, Southwest, West, Northwest directions. Another way of saying this is that the objects are 8-connected. For simplicity, sometimes people look at 4-pixel neighbourhoods, where you just look at the North, East, South and West directions. This woudl mean that the objects are 4-connected.
The output of bwlabel will give you an integer map where each object is assigned a unique ID. As such, the output of bwlabel would look something like this:
0 0 0 0 0 3 3 3 0 0
0 1 0 1 0 0 3 3 0 0
0 1 1 1 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 4
0 0 0 0 0 0 0 0 4 4
0 0 2 2 2 2 0 0 4 4
Because MATLAB processes things in column major, that's why the labelling is how you see above. As such, bwlabel gives you the membership of each pixel. This tells you where each pixel belongs to if it falls on an object. 0 in this map corresponds to the background. To call bwlabel, you can do:
L = bwlabel(img);
img would be the binary image that you supply to the function and L is the integer map I just talked about. Additionally, you can provide 2 outputs to bwlabel, and the second parameter tells you how many objects exist in the image. As such:
[L, num] = bwlabel(img);
With our above example, num would be 4. As another method of invocation, you can specify the connected pixel neighbourhoods you would examine, and so you can do this:
[L, num] = bwlabel(img, N);
N would be the pixel neighbourhood you want to examine (i.e. 4 or 8).
regionprops
regionprops is a very useful function that I use daily. regionprops measures a variety of image quantities and features in a black and white image. Specifically, given a black and white image it automatically determines the properties of each contiguous white region that is 8-connected. One of these particular properties is the centroid. This is also the centre of mass. You can think of this as the "middle" of the object. This would be the (x,y) locations of where the middle of each object is located. As such, the Centroid for regionprops works such that for each object that is seen in your image, this would calculate the centre of mass for the object and the output of regionprops would return a structure where each element of this structure would tell you what the centroid is for each of the objects in your black and white image. Centroid is just one of the properties. There are other useful features as well, but I'm assuming you don't want to do this. To call regionprops, you would do this:
s = regionprops(img, 'Centroid');
The above code will calculate the centroids of each of your objects in the image. You can specify additional flags to regionprops to specify each feature that you want. I do highly encourage that you take a look at all of the possible features that regionprops can calculate, as there are many that are useful in a variety of different applications and situations.
Also, by omitting any flags as input into the function, you would calculate all of the features in your image by default. Therefore, if we were to declare the image that we have seen above in MATLAB, this is what would happen after I run regionprops. After, let's calculate what the centroids are:
img = logical(...
[0 0 0 0 0 1 1 1 0 0;
0 1 0 1 0 0 1 1 0 0;
0 1 1 1 0 0 0 0 0 0;
0 0 0 0 0 0 0 0 0 1;
0 0 0 0 0 0 0 0 1 1;
0 0 1 1 1 1 0 0 1 1]);
s = regionprops(img, 'Centroid');
... and finally when we display the centroids:
>> disp(cat(1,s.Centroid))
3.0000 2.6000
4.5000 6.0000
7.2000 1.4000
9.6000 5.2000
As such, the first centroid is located at (x,y) = (3, 2.6), the next centroid is located at (x,y) = (4.5, 6) and so on. Take special note that the x co-ordinate is the column while the y co-ordinate is the row.
Hope this is clear!

Form a Matrix From a Large Text File Quickly

Hi I am struggling with reading data from a file quickly enough. ( Currently left for 4hrs, then crashed) must be a simpler way.
The text file looks similar like this:
From To
1 5
3 2
2 1
4 3
From this I want to form a matrix so that there is a 1 in the according [m,n]
The current code is:
function [z] = reed (A)
[m,n]=size(A);
i=1;
while (i <= n)
z(A(1,i),A(2,i))=1;
i=i+1;
end
Which output the following matrix, z:
z =
0 0 0 0 1
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
My actual file has 280,000,000 links to and from, this code is too slow for this size file. Does anybody know a much faster was to do this in matlab?
thanks
You can do something along the lines of the following:
>> A = zeros(4,5);
>> B = importdata('testcase.txt');
>> A(sub2ind(size(A),B.data(:,1),B.data(:,2))) = 1;
My test case, 'testcase.txt' contains your sample data:
From To
1 5
3 2
2 1
4 3
The result would be:
>> A
A =
0 0 0 0 1
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
EDIT - 1
After taking a look at your data, it seems that even if you modify this code appropriately, you may not have enough memory to execute it as the matrix A would become too large.
As such, you can use sparse matrices to achieve the same as given below:
>> B = importdata('web-Stanford.txt');
>> A = sparse(B.data(:,1),B.data(:,2),1,max(max(B.data)),max(max(B.data)));
This would be the approach I'd recommend as your A matrix will have a size of [281903,281903] which would usually be too large to handle due to memory constraints. A sparse matrix on the other hand, maintains only those matrix entries which are non-zero, thus saving on a lot of space. In most cases, you can use sparse matrices more-or-less as you use normal matrices.
More information about the sparse command is given here.
EDIT - 2
I'm not sure why it isn't working for you. Here's a screenshot of how I did it in case that helps:
EDIT - 3
It seems that you're getting a double matrix in B while I'm getting a struct. I'm not sure why this is happening; I can only speculate that you deleted the header lines from the input file before you used importdata.
Basically it's just that my B.data is the same as your B. As such, you should be able to use the following instead:
>> A = sparse(B(:,1),B(:,2),1,max(max(B)),max(max(B)));

Resources