計算攝影學
Computational Photography
| 節 | 週三 |
|---|---|
5 13:20–14:10 | 計算攝影學 EE527 3 節連堂 |
6 14:20–15:10 | |
7 15:30–16:20 |
* 根據陽明交大上課時間表所列
Digital imaging represents one of the most precise and highest bandwidth forms of sensing available today. The adoption of digital imaging is further encouraged by increasingly more powerful image processing techniques (collectively known as “Computational Photography”). As examples, drones, self-driving car, social media face-recognition, food and package industry etc. all rely on machine vision/imaging as their primary form of sensor. This class prepares students for a future research and engineering career that is inevitably tied to computational photography. Course objectives: • Knowledge of digital camera and lens. • Common image-processing techniques. • Calibration techniques for technical imaging (color, size, 3D calibration etc.). • Examples of unconventional computational photography techniques. • Applications: 3D measurement, machine vision, drone-imaging etc. TL,DR: "How to use camera to measure things scientifically."
• Engineering mathematics. • Matlab/Python coding or equivalent. • We'll use smartphone to take a lot of raw data. • Course lecture and assignment in English.
1.5-2hr lecture per week. Followed by ~1hr project work, demo and Q&A time. Classes will strive to be interactive, and adjusted around a few core topics based on evolving students needs and interests.
20-25% each across 4-5 hands-on, mini-projects. No exam.
- Phase 1: The Measurement Foundation (Converting photons to high-quality digital info)
- Phase 2: Spatial Mapping & Reconstruction (Mapping 2D image plane to 3D physical world)
- Phase 3: High-Dimensional Sensing (Imaging beyond the standard framework)
- Phase 4: ME Integration (Measurements for fluid and solids)
| 週次 | 主題 |
|---|---|
| 第 1 週 | (1.1) Intro to Sensor and Camera ------------------ Understanding the sensor, its performance, and the fidelity of saved data (e.g. SNR, bit-depth) |
| 第 2 週 | (1.2) Geometric Optics ------------------ Understanding the "eye" of the imaging system. Lens basics, and analytical tools in thin lens equation, RTM analysis etc. |
| 第 3 週 | (1.3) Image Pre-Processing ------------------ The "dark art" of image scientist: conditioning and standardizing acquired images for further quantitative analysis. E.g. background-subtraction, flat-fielding, blur/sharpening kernels etc. |
| 第 4 週 | (1.4) Feature Extraction from Image ------------------ Moving from pixels to measurements. Binarization, edge-detection, regionprops, centroiding, circularity etc. |
| 第 5 週 | (2.1) 2D Calibration ------------------ Homographies, checkboard patterns, lens distortion models. Image dewarp. |
| 第 6 週 | (2.2) 3D Calibration ------------------ Pinhole model, intrinsic & extrinsic parameters. Two camera-calibration. |
| 第 7 週 | (Dept Anniversary skipped) |
| 第 8 週 | (2.3) Stereo & Depth ------------------ Epipolar geometry, disparity mapping (point cloud), triangulation for 3D. |
| 第 9 週 | (2.4) Tomographic Imaging ------------------ Radon transform, back-projection, reconstructing 3D volumes from 2D projections. |
| 第 10 週 | (3.1) Light-Field Plenoptic Camera |
| 第 11 週 | (3.2) Light-Field: Camera Array |
| 第 12 週 | (3.3) Light-Field: Camera Array (continued) |
| 第 13 週 | (3.4) Event-Based: Theory ------------------ Change-detection, latency, temporal resolution. xyt-visualization. |
| 第 14 週 | (3.4) Event-Based: Processing ------------------ Asynchronous feature-tracking, noise filtering etc. |
| 第 15 週 | (4.1) Applied Metrology: For Fluid Flow ------------------ PIV, PTV |
| 第 16 週 | (No Final Exam) |
Materials are provided. Basic course contents as listed. Subject to changes.
- 聯絡方式
- tanzu@nycu.edu.tw