ok, I see, on the product page.
That’s mostly, but not entirely, wrong.
Internally the raw data we get from the sensor is about that (IIRC, it’s actually 640x480), but it’s in a format that’s not very useful for the color detecting algorithm. That image is converted and scaled down to 316x212 for processing, that’s also the coordinate system used for sending object information back to the V5.
see this kb article.