Sliding window over a 2D vector, implementation of Marching Squares algorithm

Question

I have not used C++ in quite a while and looking for advice/feedback, the code below implements the Marching Squares Algorithm which is detailed here. While the implementation appears correct and gives the same result as the example on the Wikipedia page, I think that some of the code could be improved and I'm looking for general feedback on how. In particular,

• how I handle vectors
• the 2x2 sliding matrix that moves across the test data
• where I OR in a value

---

#include <iostream>
#include <stdio.h>
#include <tchar.h>
#include <vector>
using std::vector;
void ApplyThreshold(vector<vector<int>> &Matrix2D, int threshold);
vector<vector<int>> CreateContourMatrix(vector<vector<int>> &pMatrix);
int main()
{
 const int cutOff = 2;//The value on which to base the creation of contours
 vector<vector<int>> matrix;
 //Setup test data
 matrix.push_back(vector<int> { 1, 1, 1, 1, 1 });
 matrix.push_back(vector<int> {1, 2, 3, 2, 1, });
 matrix.push_back(vector<int> {1, 3, 3, 3, 1, });
 matrix.push_back(vector<int> {1, 2, 3, 2, 1, });
 matrix.push_back(vector<int> {1, 1, 1, 1, 1, });
 vector<vector<int>> pMatrix = matrix;
 ApplyThreshold(pMatrix, cutOff);//Establishthe contours
 auto contourMatrix = CreateContourMatrix(pMatrix);//Build contour shapes
 //Print the result
 for (auto contour_row : contourMatrix)
 {
 for (int element : contour_row)
 {
 std::cout << element << '\t';
 }
 std::cout << std::endl;
 }
 return 0;
}
/*
 * Move a 2x2 window across a 2d matrix, calculating contour shapes
 */
vector<vector<int>> CreateContourMatrix(vector<vector<int>> &pMatrix)
{
 vector<vector<int>> contourMatrix;
 //vector<int> rows[5];
 for (int i =0;i<pMatrix.size() - 1;i++)
 {
 vector<int> row;//Should this allocation be moved out of the for loop?
 for (int j=0;j<pMatrix[0].size() - 1;j++)//Assume that all rows of the image are the same size
 {
 int shape = 0;
 if (pMatrix[i][j] == 1)
 shape = shape | 8;
 if (pMatrix[i][j + 1] == 1)
 shape = shape | 4;
 if (pMatrix[i+1][j] == 1)
 shape = shape | 1;
 if (pMatrix[i+1][j+1] == 1)
 shape = shape | 2;
 row.push_back(shape);
 }
 contourMatrix.push_back(row);
 }
 return contourMatrix;
}
void ApplyThreshold(vector<vector<int>> &Matrix2D, int threshhold)
{
 for (auto rowVector : Matrix2D)
 {
 for (auto vectorValue : rowVector)
 {
 if (vectorValue >= threshhold)
 vectorValue = 1;
 else
 vectorValue = 0;
 }
 }
}

This isn't supposed to be a complete implementation, i.e. I don't deal with ambiguous cases or actually create an image of the contours. Any advice appreciated.

Answer 1

A couple miscellaneous things.

I don't think you need the vector<int> for the initialization. eg:

matrix.push_back({ 1, 1, 1, 1, 1, });

You can use references in the for loops, just for good measure. IMO using the type instead of auto makes it easier to read when it's a simple type too.

for (auto& contour_row : contourMatrix)
{
 for (int& element : contour_row)
 {
 std::cout << element << '\t';
 }
 std::cout << std::endl;
}

In CreateContourMatrix() the row vector can come out of the for loop. Use .clear() or .fill() on it inside the loop, then you aren't reallocating.