1 JianJiun Ding National Taiwan University 723 531

1 丁建均 (Jian-Jiun Ding) National Taiwan University 辦公室：明達館 723室， 實驗室：明達館 531室 聯絡電話： (02)33669652 組別： 通信組，資網組 Major： Digital Signal Processing Digital Image Processing E-mail: jjding@ntu. edu. tw

2 Research Fields (1) Computer Vision (page 3) (2) Image Compression (page 4) (3) Segmentation (page 12) (4) Face Detection (page 21) (5) Character Verification (page 27) (6) Time-Frequency Analysis (page 31) (7) Music Signal Analysis (page 42) (8) Prediction and Optimization (page 47) (9) ECG Signal Analysis (page 49) (10) Other Topics (Page 54) 專題研究相關規定 (page 58)

5 2. Image Compression 目標：完成 JPEG 程式 Image 8× 8 (切成blocks) 4: 2: 2 or 4: 2: 0 8 8 DCT 量子化表 AC係數 Zigzag Scan DC係數 差分 編碼 Huffman Coding JPEG file Huffman Coding 檔頭

6 2. Image Compression JPEG 可將原圖壓縮至原來的 1/10 (for gray-level images) 或 1/20 (for color images) 然而人眼不易察覺壓縮前和壓縮後影像之間的差別 原圖 用 JPEG 壓成 1/30

7 2. Image Compression 學期後幾週的進階研究 Entropy and Compression Arithmetic Coding Adaptive Arithmetic Coding

8 2. Image Compression 學期後幾週的進階研究 Entropy and Compression Arithmetic Coding Adaptive Arithmetic Coding

10 2. Image Compression New Method: Edge-Based Segmentation and Compression 和小時候畫圖的方法類似

2. Image Compression 折衷的方法： 既不按照 8 8 的方格來做切割，也不完全按照物體的形狀 Triangular and Trapezoid (梯形) Block Segmentation 11

3. Segmentation Important for compression biomedical engineering object identification 要學習的項目： (1) 影像處理的基本技巧 (2) Image segmentation 程式的編寫 (3) Image segmentation 的應用 (4) 查資料和做研究的方法和技巧 12

3. Segmentation 13

3. Segmentation 16 Proposed Method

3. Segmentation 17

3. Segmentation 18 Proposed Method

19 3. Segmentation 學習 image segmentation 主題的學長姐們第二個學期之後的研究主題： (1) 更精準的分割技術 (2) Application for Medical Images (3) Other Applications 未受過傷的老鼠肌肉纖維

20 3. Segmentation 大腦核磁共振影像 (Brain MRI Image) (a) Brain MRI Image (b) White Matter (白質) (c) Gray Matter (灰質) (d) 腦髓液蛋白，頭蓋骨

4. Face Detection 22

4. Face Detection 最簡單的方法： matched filter 但技術上的問題頗多………. scaling shadow rotation partially distortion 目前較常用的方法： Feature Extraction + Machine Learning 臉有哪些特徵？ 23

4. Face Detection 24

25 4. Face Detection It includes the techniques of -- feature extraction, -- image processing, -- graphics -- machine learning

27 5. Character Verification Category “A” Category “B” Classification Clustering

5. Character Verification 29

6. Time-Frequency Analysis Fourier transform (FT) Time-Domain Frequency Domain Some things make the FT not practical: (1) Only the case where t 0 t t 1 is interested. (2) Not all the signals are suitable for analyzing in the frequency domain. It is hard to analyze the signal whose instantaneous frequency varies with time. 32

6. Time-Frequency Analysis Example: x(t) = cos( t) when t < 10, x(t) = cos(3 t) when 10 t < 20, x(t) = cos(2 t) when t 20 (FM signal) 33

34 6. Time-Frequency Analysis Using Time-Frequency analysis x(t) = cos( t) when t < 10, x(t) = cos(2 t) when t 20 x(t) = cos(3 t) when 10 t < 20, (FM signal) f -axis t -axis Left：using Gray level to represent the amplitude of X(t, f) Right：slicing along t = 15 t -axis

35 6. Time-Frequency Analysis Several Time-Frequency Distribution Short-Time Fourier Transform (STFT) with Rectangular Mask Gabor Transform avoid cross-term less clarity Wigner Distribution Function with cross-term high clarity Gabor-Wigner Transform (Proposed) avoid cross-term high clarity Hilbert-Huang Transform

6. Time-Frequency Analysis Applications of Time-Frequency Analysis (1) Finding Instantaneous Frequency (2) Music Signal Analysis (3) Sampling Theory (4) Modulation and Multiplexing (11) Signal Identification (12) Acoustics (13) Biomedical Engineering (14) Spread Spectrum Analysis (5) Filter Design (6) Random Process Analysis (7) Signal Decomposition (8) Electromagnetic Wave Propagation (9) Optics (10) Radar System Analysis (15) System Modeling (16) Image Processing (17) Economic Data Analysis (18) Signal Representation (19) Data Compression (20) Seismology (21) Geology 36

6. Time-Frequency Analysis Conventional Sampling Theory Nyquist Criterion New Sampling Theory (1) t can vary with time (2) Number of sampling points == Area of time frequency distribution 37

6. Time-Frequency Analysis An Example of the Hilbert-Huang Transform Envelopes 39

6. Time-Frequency Analysis IMF 1 IMF 2 x 0(t) 40

43 7. Music Signal Analysis Using the time-frequency analysis 聲音檔：http: //djj. ee. ntu. edu. tw/Chord. wav So Mi Do La Fa Re

7. Music Signal Analysis 聲音檔：http: //djj. ee. ntu. edu. tw/air. mp 3 time-frequency analysis 44

47 8. Prediction and Optimization 研究使用信號處理技術的 prediction (預測) 演算法 以及研究最佳化 (optimization) 的演算法 Prediction Techniques are important for signal analysis (信 號 分 析 )， motion prediction (動 作 預 測 )， whether forecasting (天 氣 預 測 )， economical signal prediction (經濟信號預測) ，risk accessment (風險評 估)，and other applications. Optimization is important for signal analysis (信號分析), modeling (模型 建構), and machine learning (機器學習)。

48 8. Prediction and Optimization Sub-topics about Prediction and Optimization (1) PCA and SVD (2) Particle and Kalman Filters (3) Regression and Linear Prediction Model (4) Nonlinear Prediction Model (5) Newton’s Method, Golden Section Search, and Gradient Descent (6) Optimization for Multiple Parameters (7) L 2 Norm Optimization (8) L Norm Optimization

9. ECG (心電圖) Signal Analysis 典型心電圖 R R P T QS P QS T 50

9. ECG (心電圖) Signal Analysis (a) The Original Signal (The First ECG Signal in 9. bmp) (b) Find the Baseline (c) Subtracted by the Baseline

9. ECG (心電圖) Signal Analysis 53 Telehealth (遠距醫療) Can we perform health examination by the ibon machine in 7 -11 or at home?