for i = 2:N
A(i,i-1:i+1) = [1, -2, 1];
end
Hello, matlab is telling me that this code can be faster by using spalloc for the matrix A (which I have) but also by vectorizing this for loop. I've tried to use the following:
i = 2:N
A(i, i-1:i+1)
but the result obviously turned out to be not what I want.
How can I solve this?
Thank you!
It looks like you're trying to get a second-order difference operator, except your loop winds up missing the first row and including an extra column. The normal (sparse) difference operator is generated like this:
N = 10;
v = ones(N, 1);
A = spdiags([v -2*v v], [-1 0 1], N, N);
full(A) % for display only
You'll see:
ans =
-2 1 0 0 0 0 0 0 0 0
1 -2 1 0 0 0 0 0 0 0
0 1 -2 1 0 0 0 0 0 0
0 0 1 -2 1 0 0 0 0 0
0 0 0 1 -2 1 0 0 0 0
0 0 0 0 1 -2 1 0 0 0
0 0 0 0 0 1 -2 1 0 0
0 0 0 0 0 0 1 -2 1 0
0 0 0 0 0 0 0 1 -2 1
0 0 0 0 0 0 0 0 1 -2
If that's not quite what you want (e.g., you really don't want the first row), then it's probably faster to generate it as above and then fix it up.
Given a boolean 2D matrix (0-based index), find whether there is path from (0,0) to (x,y) and if there is one path, print the minimum no of steps needed to reach it, else print -1 if the destination is not reachable. You may move in only four direction ie up, down, left and right. The path can only be created out of a cell if its value is 1.
Example:
Input:
2
3 4
1 0 0 0 1 1 0 1 0 1 1 1
2 3
3 4
1 1 1 1 0 0 0 1 0 0 0 1
0 3
Output:
5
3
Input:
The first line of input contains an integer T denoting the no of test cases. Then T test cases follow. Each test case contains two lines . The first line of each test case contains two integers n and m denoting the size of the matrix. Then in the next line are n*m space separated values of the matrix. The following line after it contains two integers x and y denoting the index of the destination.
Output:
For each test case print in a new line the min no of steps needed to reach the destination.
Code:
bool isSafe(int currRow,int currCol,int rows,int columns,vector<bool> visited[]) {
return currRow>=0 && currRow<rows && currCol>=0 && currCol<columns && !visited[currRow][currCol];
}
int minSteps(vector<int> matrix[],int n,int m,int x,int y) {
vector<bool> visited[n];
for(int i=0;i<n;i++){
vector<bool> tmp(m);
for(int j=0;j<m;j++){
if(matrix[i][j]==0){
tmp[j]=true;
} else {
tmp[j]=false;
}
}
visited[i]=tmp;
}
queue<pair<int,int>> q;
q.push(make_pair(0,0));
visited[0][0]=true;
int minDist[n][m];
for(int i=0;i<n;i++){
for(int j=0;j<m;j++){
minDist[i][j]=INT_MAX;
}
}
minDist[0][0]=0;
static int rows[]={0,1,0,-1};
static int columns[]={1,0,-1,0};
while(!q.empty()) {
pair<int,int> p=q.front();
q.pop();
for(int i=0;i<4;i++) {
if(isSafe(p.first+rows[i],p.second+columns[i],n,m,visited)) {
visited[p.first+rows[i]][p.second+columns[i]]=true;
q.push(make_pair(p.first+rows[i],p.second+columns[i]));
if(minDist[p.first+rows[i]][p.second+columns[i]]> minDist[p.first][p.second]+1) {
minDist[p.first+rows[i]][p.second+columns[i]] = minDist[p.first][p.second]+1;
}
}
}
}
if(minDist[x][y]!=INT_MAX) {
return minDist[x][y];
}
return -1;
}
Test Case Failing
Input:
20 13
0 1 1 1 1 0 0 0 1 1 1 1 0 0 0 0 0 1 1 1 0 1 0 1 1 0 0 1 0 1 0 1 0 1 0 1 1 0 0 0 1 1 0 1 1 0 1 1 1 0 1 1 0 1 0 1 1 0 0 1 1 0 0 0 0 0 1 1 1 1 0 0 1 0 0 0 0 1 0 1 0 1 0 0 0 0 1 1 0 1 0 1 0 1 1 0 1 1 1 0 0 1 1 1 1 1 0 1 0 0 0 0 1 0 0 1 0 1 0 0 1 0 0 1 1 1 1 1 0 0 1 1 0 0 0 1 1 0 1 0 0 1 0 1 0 1 0 0 0 0 1 1 1 1 0 0 0 1 1 1 0 1 0 0 1 0 1 1 1 0 1 1 0 1 1 0 0 1 0 1 0 1 0 1 0 1 1 1 0 1 0 0 0 0 0 1 0 0 0 1 0 1 1 0 1 0 0 1 1 1 0 1 0 1 0 1 0 0 0 0 1 0 1 1 0 1 0 0 1 1 0 0 0 1 0 1 1 1 1 0 0 1 0 0 0 0 1 0 0 1 1 1 1 0 0 1 1 1 0 1
6 3
Its Correct output is:
-1
And Your Code's output is:
13
Algorithm:
1. Traverse the 2 d array from source using BFS.
2. Maintain 2 2d arrays visited and minDist.
3. Initialize values of visited as true whose value in array is 0 and rest as false; Initialize minDist to INT_MAX.
4. While traversing, validate if its a valid point using isSafen where it is checked if visited is false and point lies within 2d array size limits.
5. If point is safe, make visited for the point as true and push it in the queue.
6. Finlly check if mindist for the point is greater than its parent minDist + 1 ; Update accordingly.
But i am getting wrong answer; attached failing test case. Can someone explain where my algo is going wrong ?
I have the missed below corner case:
If matrix[0,0] == 0
return -1
Now, algorithm passes all test cases.
I tried to trace the Perceptron algorithm for logical "Or" with binary input (0,1) and binary output (0,1). But, it seems like that it doesn't work!
Here is my try:
x1 x2 w1 w2 bias t y
1 1 0 0 0 1 0 Update
1 0 1 1 1 1 1 OK
0 1 1 1 1 1 1 OK
0 0 1 1 1 0 1 Update
1 1 1 1 1 1 1 OK
1 0 1 1 1 1 1 OK
0 1 1 1 1 1 1 OK
0 0 1 1 1 0 1 Update (but as before no updates occur)
My update rules are:
Wi = Wi + xi*ti
Bi = Bi + ti
It seems my update rule was very simple. The exact update rule must be:
Wi = Wi + xi*(ti - yi)
Bi = Bi + (ti - yi)
This change causes to have a -1 for updating b when both x1 and x2 are zero:
x1 x2 w1 w2 bias t y t-y
1 1 0 0 0 1 0 1 Update
1 0 1 1 1 1 1 0 OK
0 1 1 1 1 1 1 0 OK
0 0 1 1 1 0 1 -1 Update
1 1 1 1 0 1 1 0 OK
1 0 1 1 0 1 1 0 OK
0 1 1 1 0 1 1 0 OK
0 0 1 1 0 0 1 0 OK
I am trying to find islands of numbers in a matrix.
By an island, I mean a rectangular area where ones are connected with each other either horizontally, vertically or diagonally including the boundary layer of zeros
Suppose I have this matrix:
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1
0 0 0 1 1 1 0 1 1 0 0 0 1 1 1 1 0
0 0 0 0 0 0 1 0 1 0 0 0 0 1 1 1 1
0 0 0 1 0 1 0 1 1 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 1 0 1 1 1 0 0 0 0 0 0 0
0 0 1 0 1 1 1 1 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
By boundary layer, I mean row 2 and 7, and column 3 and 10 for island#1.
This is shown below:
I want the row and column indices of the islands. So for the above matrix, the desired output is:
isl{1}= {[2 3 4 5 6 7]; % row indices of island#1
[3 4 5 6 7 8 9 10]} % column indices of island#1
isl{2}= {[2 3 4 5 6 7]; % row indices of island#2
[12 13 14 15 16 17]}; % column indices of island#2
isl{3} ={[9 10 11 12]; % row indices of island#3
[2 3 4 5 6 7 8 9 10 11];} % column indices of island#3
It doesn't matter which island is detected first.
While I know that the [r,c] = find(matrix) function can give the row and column indices of ones but I have no clues on how to detect the connected ones since they can be connected in horizontal, vertical and diagonal order.
Any ideas on how to deal with this problem?
You should look at the BoundingBox and ConvexHull stats returned by regionprops:
a = imread('circlesBrightDark.png');
bw = a < 100;
s = regionprops('table',bw,'BoundingBox','ConvexHull')
https://www.mathworks.com/help/images/ref/regionprops.html
Finding the connected components and their bounding boxes is the easy part. The more difficult part is merging the bounding boxes into islands.
Bounding Boxes
First the easy part.
function bBoxes = getIslandBoxes(lMap)
% find bounding box of each candidate island
% lMap is a logical matrix containing zero or more connected components
bw = bwlabel(lMap); % label connected components in logical matrix
bBoxes = struct2cell(regionprops(bw, 'BoundingBox')); % get bounding boxes
bBoxes = cellfun(#round, bBoxes, 'UniformOutput', false); % round values
end
The values are rounded because the bounding boxes returned by regionprops lies outside its respective component on the grid lines rather than the cell center, and we need integer values to use as subscripts into the matrix. For example, a component that looks like this:
0 0 0
0 1 0
0 0 0
will have a bounding box of
[ 1.5000 1.5000 1.0000 1.0000 ]
which we round to
[ 2 2 1 1]
Merging
Now the hard part. First, the merge condition:
We merge bounding box b2 into bounding box b1 if b2 and the island of b1 (including the boundary layer) have a non-null intersection.
This condition ensures that bounding boxes are merged when one component is wholly or partially inside the bounding box of another, but it also catches the edge cases when a bounding box is within the zero boundary of another. Once all of the bounding boxes are merged, they are guaranteed to have a boundary of all zeros (or border the edge of the matrix), otherwise the nonzero value in its boundary would have been merged.
Since merging involves deleting the merged bounding box, the loops are done backwards so that we don't end up indexing non-existent array elements.
Unfortunately, making one pass through the array comparing each element to all the others is insufficient to catch all cases. To signal that all of the possible bounding boxes have been merged into islands, we use a flag called anyMerged and loop until we get through one complete iteration without merging anything.
function mBoxes = mergeBoxes(bBoxes)
% find bounding boxes that intersect, and merge them
mBoxes = bBoxes;
% merge bounding boxes that overlap
anyMerged = true; % flag to show when we've finished
while (anyMerged)
anyMerged = false; % no boxes merged on this iteration so far...
for box1 = numel(mBoxes):-1:2
for box2 = box1-1:-1:1
% if intersection between bounding boxes is > 0, merge
% the size of box1 is increased b y 1 on all sides...
% this is so that components that lie within the borders
% of another component, but not inside the bounding box,
% are merged
if (rectint(mBoxes{box1} + [-1 -1 2 2], mBoxes{box2}) > 0)
coords1 = rect2corners(mBoxes{box1});
coords2 = rect2corners(mBoxes{box2});
minX = min(coords1(1), coords2(1));
minY = min(coords1(2), coords2(2));
maxX = max(coords1(3), coords2(3));
maxY = max(coords1(4), coords2(4));
mBoxes{box2} = [minX, minY, maxX-minX+1, maxY-minY+1]; % merge
mBoxes(box1) = []; % delete redundant bounding box
anyMerged = true; % bounding boxes merged: loop again
break;
end
end
end
end
end
The merge function uses a small utility function that converts rectangles with the format [x y width height] to a vector of subscripts for the top-left, bottom-right corners [x1 y1 x2 y2]. (This was actually used in another function to check that an island had a zero border, but as discussed above, this check is unnecessary.)
function corners = rect2corners(rect)
% change from rect = x, y, width, height
% to corners = x1, y1, x2, y2
corners = [rect(1), ...
rect(2), ...
rect(1) + rect(3) - 1, ...
rect(2) + rect(4) - 1];
end
Output Formatting and Driver Function
The return value from mergeBoxes is a cell array of rectangle objects. If you find this format useful, you can stop here, but it's easy to get to the format requested with ranges of rows and columns for each island:
function rRanges = rect2range(bBoxes, mSize)
% convert rect = x, y, width, height to
% range = y:y+height-1; x:x+width-1
% and expand range by 1 in all 4 directions to include zero border,
% making sure to stay within borders of original matrix
rangeFun = #(rect) {max(rect(2)-1,1):min(rect(2)+rect(4),mSize(1));...
max(rect(1)-1,1):min(rect(1)+rect(3),mSize(2))};
rRanges = cellfun(rangeFun, bBoxes, 'UniformOutput', false);
end
All that's left is a main function to tie all of the others together and we're done.
function theIslands = getIslandRects(m)
% get rectangle around each component in map
lMap = logical(m);
% get the bounding boxes of candidate islands
bBoxes = getIslandBoxes(lMap);
% merge bounding boxes that overlap
bBoxes = mergeBoxes(bBoxes);
% convert bounding boxes to row/column ranges
theIslands = rect2range(bBoxes, size(lMap));
end
Here's a run using the sample matrix given in the question:
M =
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1
0 0 0 1 1 1 0 1 1 0 0 0 1 1 1 1 0
0 0 0 0 0 0 1 0 1 0 0 0 0 1 1 1 1
0 0 0 1 0 1 0 1 1 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 1 0 1 1 1 0 0 0 0 0 0 0
0 0 1 0 1 1 1 1 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
>> getIslandRects(M)
ans =
{
[1,1] =
{
[1,1] =
9 10 11 12
[2,1] =
2 3 4 5 6 7 8 9 10 11
}
[1,2] =
{
[1,1] =
2 3 4 5 6 7
[2,1] =
3 4 5 6 7 8 9 10
}
[1,3] =
{
[1,1] =
2 3 4 5 6 7
[2,1] =
12 13 14 15 16 17
}
}
Quite easy!
Just use bwboundaries to get the boundaries of each of the blobs. you can then just get the min and max in each x and y direction of each boundary to build your box.
Use image dilation and regionprops
mat = [...
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 1;
0 0 0 1 1 1 0 1 1 0 0 0 1 1 1 1 0;
0 0 0 0 0 0 1 0 1 0 0 0 0 1 1 1 1;
0 0 0 1 0 1 0 1 1 0 0 0 1 0 0 0 0;
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0;
0 0 0 1 0 1 0 1 1 1 0 0 0 0 0 0 0;
0 0 1 0 1 1 1 1 1 0 0 0 0 0 0 0 0;
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0];
mat=logical(mat);
dil_mat=imdilate(mat,true(2,2)); %here we make bridges to 1 px away ones
l_mat=bwlabel(dil_mat,8);
bb = regionprops(l_mat,'BoundingBox');
bb = struct2cell(bb); bb = cellfun(#(x) fix(x), bb, 'un',0);
isl = cellfun(#(x) {max(1,x(2)):min(x(2)+x(4),size(mat,1)),...
max(1,x(1)):min(x(1)+x(3),size(mat,2))},bb,'un',0);
I'm working on Brushfire algorithm and I need to make a loop which will scan through the matrix and find the adjacent zeros with ones and convert "1" to "2". Assume that I have a matrix 5 by 5:
0 0 0 0 0
0 1 1 1 1
0 0 1 1 1
0 0 1 1 1
0 0 1 1 1
Can I somehow make it:
0 0 0 0 0
0 2 2 2 2
0 0 2 1 1
0 0 2 1 1
0 0 2 1 1
Thank you
With the image processing toolbox, the algorithm would be:
A = [0 0 0 0 0
0 1 1 1 1
0 0 1 1 1
0 0 1 1 1
0 0 1 1 1];
B = A;
%# set pixels at border between 0 and 1 to 2
B(imdilate(~A,true(3)) & A>0) = 2;
You do it with 2D-convolution, using the standard function conv2. Denoting your matrix as X,
mask = [0 1 0; 1 1 1; 0 1 0]; %// or [1 1 1; 1 1 1; 1 1 1] to include diagonal adjacency
X(conv2(double(~X), mask, 'same') & X) = 2;