Why not we use original image instead decoded image to P-frame?

I'm trying to want to know the P-frame at mpeg. I have a query about reference image.

Why not we use original image instead decoded image to make P-frame?

Solution

I-frame, B-frame and P-frame allows to compress the video.

Indeed, in a video you have a lot of redundant information. Think about a car moving across the screen: all the pixels in the background do not change from a picture to another, only those around the car are "moving". With the I-B-P frame truck, you give the code of the background and then, you just signalling slight changes (the car moving) through vectors.

This way you have to carry less information than if you have to repeat the entire picture each time.