Why Panning Looks Worse Than Anything Else On Screen

Nothing exposes the weaknesses of a video system quite like a slow pan across a detailed scene. Three separate limitations surface at the same moment.
Every pixel changes at once
Compression saves most of its data by describing what stayed the same between frames. A pan moves the entire picture, so nothing stays where it was.
Motion compensation helps, since the whole frame shifts by a consistent amount and the encoder can describe that displacement compactly. Uniform movement is the easiest kind to match.
The trouble is that new content enters at the leading edge every frame, and that content has never been seen before. It must be described from scratch while the budget is already stretched.
Judder comes from the frame rate itself
During a pan, a fixed object's image steps sideways by some distance each frame. If that distance is large, the eye perceives discrete jumps rather than continuous movement.
Cinematographers have long-standing rules limiting how fast a pan may be at a given frame rate, expressed as the time a subject should take to cross the frame.
Exceeding the limit produces strobing that no amount of bitrate corrects. It is a sampling problem, not a compression one.
The eye tracks and the display does not
A viewer following a moving subject moves their eyes smoothly while the display holds each frame still for its full duration. The image slides across the retina during that hold.
The result is a perceived blur that the video does not contain, produced entirely inside the viewer's visual system. Higher frame rates shorten each hold and reduce it.
This is why panning can look softer on a large screen viewed closely. The same angular error covers more of the visual field.
Refresh mismatches add a second stutter
Displays refresh at a fixed rate that is often not a whole multiple of the content rate. Frames are then repeated unevenly to fill the schedule.
Uneven repetition adds its own rhythmic hitch on top of any judder in the source. During a pan the two combine into something clearly wrong.
Displays that can adapt their refresh rate to the content eliminate this layer. The improvement on a slow pan is usually obvious.
Handheld movement multiplies every problem
A pan on a tripod moves consistently, which motion matching handles well. Handheld movement adds unpredictable vertical and rotational components.
Rotation and scaling are poorly described by simple displacement, so the encoder falls back to expensive new picture data across the whole frame.
Stabilization applied afterward crops and warps the image, which changes the geometry again and can leave the picture soft. Recording steadily costs far less than fixing it later.



