OpenCV stores images in a data structure called IplImage. They provide methods for rendering it to the screen, but if you want to use OpenGL instead (which should be faster and gives you more flexibility), I wrote the following code. It should be a little faster than loading the IplImage into a texture first, and then drawing it (if you have to do this every frame).
format = GL_LUMINANCE;
format = GL_LUMINANCE_ALPHA;
format = GL_BGR;
glDrawPixels(image->width, image->height, format, GL_UNSIGNED_BYTE, image->imageData);
It should be pretty straight forward to use. Handles IplImages with 1 or 3 layers. Should work with 2 layers too, but I haven't needed or found an example of this.