Page 32 -

6 Chapter 1 ■ Practical Aspects of a Vision System

double in type. The function cvGet2D does this; getting the pixel value at i,j
for the image above is simply
p = cvGet2D (img, i, j);

The variable p is of type CvScalar,which is

struct CvScalar
{
double val[4];
}
If the pixel has only a single value (i.e., grey), then p.val[0] is that value. If it
is RGB, then the color components of the pixel are as follows:
Blue is p.val[0]
Green is p.val[1]

Red is p.val[2]
Modifying the pixel value is done as follows:

p.val[0] = 0; // Blue
p.val[1] = 255; // Green
p.val[2] = 255; // Red
cvSet2D(img,i,j,p); // Set the (i,j) pixel to yellow
This is referred to as indirect access in OpenCV documentation and is slower
than other means of accessing pixels. It is, on the other hand, clean and clear.

1.2.2 Reading and Writing Images
The basic function for image input has already been seen; cvLoadImage reads
an image from a file, given a path name to that file. It can read images in JPEG,
BMP, PNM, PNG, and TIF formats, and does so automatically, without the
need to specify the file type. This is determined from the data on the file itself.
Once read, a pointer to an IplImage structure is returned that will by default
be forced into a 3-channel RGB form, such as has been described previously.
So, the call
img = cvLoadImage (filename);

returns an IplImage* value that is an RGB image, unless the file name indicated
by the string variable filename can’t be read, in which case the function returns
0 (null). A second parameter can be used to change the default return image.
The call
img = cvLoadImage (filename, f);