I hope this will grow into a useful resource for other people using OpenCV in C++.

 

Copy an image into a region of interest (ROI) of a larger image

…assuming the same image type:

imageSmall.copyTo(imageBigger(Rect(iTargetX,
                                   iTargetY, 
                                   imageSmall.size().width, 
                                   imageSmall.size().height)));

 

Copy a ROI from one image into a new image

…given an existing image and a Rect describing an ROI:

Mat imageCrop;
imageSource(rectROI).copyTo(imageCrop);

 

Convert a Mat into an IplImage

…NOTE – this does not make a copy – it just creates a new header over the existing data – if the Mat or the IplImage are changed, the other will also change. To make a separate copy, do a imageMat.copyTo(dst) first.

Mat imageMat;
IplImage *imageIplImage;
imageMat = imload(...);

imageIplImage = new IplImage(imageMat);

 

Convert an IplImage into a Mat

…NOTE – haven’t yet found out whether this makes a copy of the data, or just a new header.

Mat image;
IplImage *imageIplImage;
imageMat = Mat(imageIplImage);
mageMat);

 

Decode a JPG which is already in memory

If the JPG is already loaded into memory, but not decoded (i.e. it’s just been byte-for-byte read from disk), it can be decoded as follows using imdecode. First it has to be loaded into a Matrix.

void *pImageData;     // Set this to point to the 'file' in memory
long lImageDataSize;  // Number of bytes

Mat imageDecoded = cv::imdecode(cv::Mat(1, 
                                        lImageDataSize, 
                                        CV_8UC1, 
                                        pImageData), 
                                CV_LOAD_IMAGE_UNCHANGED);

 

Encode a Mat into a JPG (or other image format) to a memory buffer instead of file

Mat imageSourceImage;
vector vectorImageEncoded;
void *pEncodedData;

imencode(".jpg", imageSource, vectorImageEncoded);
// To copy the data to a void * or similar...
pEncodedData = malloc(sizeof(char) * vectorImageEncoded.size());
memcpy(pEncodedData, &vectorImageEncoded[0], vectorImageEncoded.size());

How to fill an image with a single colour

image.setTo(cv::Scalar(iBlue,iGreen,iRed));

How to resize an image

imageSource.resize(imageSource, imageDisplay, Size(DISPLAY_COLS, DISPLAY_ROWS), INTER_NEAREST);

How to access pixels in an image

Note – this is NOT the quickest way of accessing pixels, especially accessing *all* pixels, or whole rows at a time – in those cases, there are ‘pointer arithmetic’ methods which are a lot more efficient. However, for ‘random’ access, this is the *easiest* way.
This example shows code getting the Blue, Green and Red channels for pixels coord iX,iY, from an RGB image (order of colour planes is BGR).

iB = imageBGR.at<cv::Vec3b>(iY,iX)[0] = 0;
iG = imageBGR.at<cv::Vec3b>(iY,iX)[1] = 0;
iR = imageBGR.at<cv::Vec3b>(iY,iX)[2] = 0;

Pixels can be set in exactly the same way:

imageRGB.at<cv::Vec3b>(iY,iX)[0] = 255; // Sets Blue channel

…etc

How to create a matrix

Mat matImage;
matImage.create(iRows, iCols, CV_32FC1);

or

matImage.create(imageOther.size(), CV_8UC1);

…etc

How to find contours

Mat matImage;   // 1 byte int, e.g. CV_8UC1
vector<vector<cv::Point> > contour;
vector<Vec4i> vectorHierarchy;
findContours(matImage, contour, vectorHierarchy, CV_RETR_LIST, CV_CHAIN_APPROX_NONE);
// NOTE - matImage WILL BE MODIFIED