If by change you mean replace, then you can use the image ROI (region of interest) functions to directly replace a rectangular region of your original image with a rectangular region from another image very efficiently.
Suppose your original image was stored in A
and you want to change a part of it (a rectangular region) using pixels from an image B
.
UPDATE: Here's the code in C
/**** C ****/
// Acquire Image A and B (here as an example, I'm reading from disk)
IplImage* A = cvLoadImage("image_A.jpg");
IplImage* B = cvLoadImage("image_B.jpg");
// Set the region-of-interest (ROI) for the two images
// such that only the ROI of A and B will be handled
cvSetImageROI(A,cvRect(200,200,128,128));
cvSetImageROI(B,cvRect(0,0,128,128));
// Copy the ROI in B to the ROI in A
cvCopy(B,A);
// Reset the ROI (now the entire image will be handled)
cvResetImageROI(A);
cvResetImageROI(B);
// Display A
cvNamedWindow("Modified A");
cvShowImage("Modified A",A);
cvWaitKey();
// Release the images
cvReleaseImage(&A);
cvReleaseImage(&B);
Using OpenCV 2.0:
// C++ //
// Images A and B have already been loaded .....
// Region in image A starting from (100,100) of width 200 and height 200
Rect RegionA(100,100,200,200);
// Region in image B starting from (50,50) of width 200 and height 200
Rect RegionB(50,50,200,200);
// No copying, just a reference to the ROI of the image
Mat A_ROI(A,RegionA);
Mar B_ROI(B,RegionB);
// Copy all the pixels in RegionB in B to RegionA to A
B.copyTo(A);