31. Perceptual Compression Display
- Status: Speculative
- Interface point: Perception or assistive interface
- Carrier: Any usable sensory or neural channel
- Transducer: Display, audio, haptic, retinal, or neural interface
- Encoding style: Compressed perceptual features
- Addressing method: Task-aware cue placement
- Expected percept: Useful awareness rather than full image reconstruction
- Related modules: Signal Encoding, Perceptual Rendering
Summary
A perceptual compression display asks: what can be removed from a visual scene
while still preserving usefulness?
It treats artificial vision as a codec problem. The system may not transmit
color, texture, or every object. It may transmit only the features needed for a
task.
The stronger version treats the user’s perception as part of the medium. The
device does not need to carry the whole image if the person can learn a stable
cue grammar and use movement, attention, and context to complete the scene.
Chain Map
scene
-> feature extraction
-> compression by task
-> cue vocabulary
-> learned perception
Compression Targets
| Target |
Useful For |
| Edges |
Object boundaries and reading support. |
| Motion |
Moving people, vehicles, and hazards. |
| Depth |
Navigation and reaching. |
| Object labels |
Recognition and task support. |
| Salience |
Attention guidance. |
| Confidence |
Knowing when the system might be wrong. |
Expansion Ideas
- Encode a room as five cues or fewer.
- Switch cue vocabulary based on user task.
- Use high detail only around gaze.
- Pair weak visual cues with spoken semantic labels.
- Let the user tune what matters.
- Compare passive cue viewing with active exploration.
- Keep uncertainty visible as part of the cue, not as a hidden machine state.
Research Questions
- Which compressed cues are most useful for each task?
- When does less information outperform more confusing information?
- How should uncertainty be represented?
- Can users become fluent in a compressed visual language?
- Which middle layer is being removed: pixel grid, frame stream, screen, or
photorealism?