Video Compression - Motion Estimation & Compensation

Try the playback to see the video in action. Then, look inside to see how we avoid storing every pixel!
How are the frames stored?
I-FRAME
420 KB Data Stored: 100%

Analysis: The Efficiency of Motion

By comparing frames, the computer realizes that most objects don't disappear—they just move. We use two main processes to exploit this:

1. Motion Estimation (Encoding)

The computer analyzes the I-Frame and searches for where those pixels went in the next frame. It outputs a Motion Vector—a mathematical "map" of the movement.

2. Motion Compensation (Decoding)

During playback, the computer takes the I-Frame and "warps" it using the stored map. By adding a tiny Residual (error correction), it perfectly reconstructs the new frame.

The H.264 Group of Pictures (GOP)

Modern H.264 video organizes frames into a GOP sequence, typically using an IBBP pattern. Unlike P-frames which only look backward, B-frames (Bi-directional) reference both previous and future frames.

I1
B2
B3
P4
B2
B3
P7
I8
Did you know? You can see the difference between high-efficiency IPB and professional All-Intra (All-I) recording formats in this technical resource by Canon (Chinese version here).