数字影像的本质：采样Digital Imaging Is Sampling
学习目标LEARNING OBJECTIVES
学完你能说清数字影像的本质——把连续的光切成数——以及噪点、色带这些东西的来路。
概念说明CONCEPTS
四个方向的切分：
- 空间切成网格：像素=空间采样点，格子外的信息永久丢失
- 时间切成瞬间：帧=时间采样点，帧间的运动靠眼睛补
- 幅度切成台阶：位深=量化级数，台阶不够渐变就变色带
- 颜色约定坐标系：色域+传递函数=数字怎么被解释成颜色
本质：用有限个数字近似连续的光；噪点、色带、锯齿、运动抖动——这些不是软件的锅，是近似本身的代价。
操作步骤STEPS
- 对照五方向卡逐一自检
- 把「噪点 / 色带 / 锯齿」各自对应到哪个切分
- 记住：近似本身就有代价
练习PRACTICE
动手任务
拿一张有渐变天空的照片，指出哪个 artefact 对应哪个「切分」。
自检清单
做完任务后逐条自查，全部能打勾才算过。
常见问题FAQ
噪点是软件的锅吗？
不是——是用有限个数字近似连续的光，本身的代价。
采样时丢掉的信息还能找回来吗？
不能——格子外的信息永久丢失；数字影像本来就是用有限个数字近似连续的光。
Learning Objectives学习目标
After this card you can say what digital imaging really is — cutting continuous light into numbers — and where noise and banding come from.
Concepts概念说明
Four directions of cutting:
- Space into a grid: pixels are spatial samples; outside the cells information is lost for good
- Time into instants: frames are temporal samples; your eye fills in between
- Amplitude into steps: bit depth = quantisation levels; too few and gradients band
- Colour by agreed coordinates: gamut + transfer function = how numbers become colour
The essence: approximating continuous light with a finite set of numbers. Noise, banding, aliasing, motion judder — not the software's fault; the price of approximation itself.
Steps操作步骤
- Self-check against the five-axis cards
- Map noise / banding / aliasing to their cut
- Remember: approximation itself has a price
Practice练习
Hands-on task
Take a photo with a gradient sky and point out which artifact maps to which "cut".
Self-check
Work through the task, then tick every line; all checked means you own this card.
FAQ常见问题
Is noise the software's fault?
No — it is the price of approximating continuous light with a finite set of numbers.
Can information lost at sampling be recovered?
No — what's outside the grid is gone for good; digital images are by nature finite numbers approximating continuous light.