The change in view with the change in angle of incidence is one the *problems* with existing mirrors. A camera could be permanently fixed to view exactly the right area, whereas mirrors need to be properly adjusted for each driver, and don't work correctly if you move your head even a little.
Plus it's video -- you can take multiple images and stitch them together (or display multiple views in a single location, as is common in side mirrors in larger vehicles) and you're not limited to the vantage point of the traditional mirror. For example, a combination of a side-rear looking camera from near the driver's position and a side-looking camera from near the back of the car -- and both could be mounted up high, rather than below the window line -- would provide better field-of-view than virtually any existing side view mirror.
Plus no reflected headlights/sun. Heck, with high camera mounting points you can significantly reduce the possibility even of shining a headlight into the camera, let alone blinding the driver.
And of course once you've put a sensor pod on the side of the car and a display in the dash, adding things like ultrasonic proximity detection become much cheaper and easier to integrate into existing driving methodologies.