Presentation is loading. Please wait.

Presentation is loading. Please wait.

XIII. Continuous WT with Discrete Coefficients

Similar presentations


Presentation on theme: "XIII. Continuous WT with Discrete Coefficients"— Presentation transcript:

1 XIII. Continuous WT with Discrete Coefficients
13-A Definition The parameters a and b are not chosen arbitrarily. For example, a = n2m and b = 2m. 註:某些文獻把這個式子稱作是 discrete wavelet transform,實際上仍然是 continuous wavelet transform 的特例  Main reason for constrain a and b to be n2m and 2m : Easy to implementation Xw(n, m) can be computed from Xw(n, m−1) by digital convolution.

2 13-B Inverse Wavelet Transform
1(t) is the dual function of (t). i.e., should be satisfied.

3 We often desire that 1(t) = (t).
The mother wavelet should satisfies

4 13-C Haar Wavelet (t) mother wavelet (wavelet function)
(t) scaling function 1 t=0 t= 0.5 t=1 t = 1 t = 0 (2t) (2t) (2t) = (t) + (t) 0 0.5 (t) = (2t) + (2t–1) (t) = (2t) – (2t–1)

5 (2mt − n) (2mt − n) t=n2-m t=(n+.5)2-m t=(n+1)2-m t=n2-m t=(n+1)2-m  Advantages of Haar wavelet (1) Simple (2) Fast algorithm (3) Orthogonal →reversible (4) Compact, real, odd (5) Vanish moment

6 Properties (1) 任何 function 都可以由 (t), (2t), (4t), (8t), (16t), ……….. 以及它們的位移所組成 (2) 任何平均為 0 的function 都可以由 (t), (2t), (4t), (8t), (16t), ……….. 所組成 換句話說……. 任何 function 都可以由 constant, (t), (2t), (4t), (8t), (16t), ……….. 所組成

7 (3) Orthogonal The dual function of (t) is (t) itself.

8 (4) 不同寬度 (也就是不同 m) 的 wavelet / scaling functions 之間會有一個關係
(t) = (2t) + (2t − 1) (t − n) = (2t − 2n) + (2t − 2n − 1) (2mt − n) = (2m+1t − 2n) + (2m+1t − 2n − 1) (t) = (2t) − (2t − 1) (t − n) = (2t − 2n) − (2t − 2n − 1) (2mt − n) = (2m+1t − 2n) − (2m+1t − 2n − 1)

9 (5) 可以用 m+1 的 coefficients 來算 m 的 coefficients

10 layer: w[n, m] w[2n, m +1] w[4n, m +2] w[4n+1, m +2] Xw[2n, m +1] −1 w[2n+1, m+1] Xw[n, m] w[4n+2, m +2] −1 Xw[2n+1, m +1] w[4n+3, m +2] −1 structure of multiresolution analysis (MRA)

11 13-D General Methods to Define the Mother Wavelet and the Scaling Function
Constraints: (a) nearly compact support (b) fast algorithm (c) real (d) vanish moment (e) orthogonal 和 continuous wavelet transform 比較: (1) compact support 放寬為 “near compact support” (2) 沒有 even, odd symmetric 的限制 (3) 由於只要是 complete and orthogonal, 必定可以 reconstruction 所以不需要 admissibility criterion 的限制 (4) 多了對 fast algorithm 的要求

12 13-E Fast Algorithm Constraints
Higher and lower resolutions 的 recursive relation 的一般化 稱作 dilation equation (t): mother wavelet, (t): scaling function 這些關係式成立,才有fast algorithms

13 If then If then

14 (Step 1) convolution (Step 2) down sampling

15  2  2 m 越大,越屬於細節

16  To satisfy , FT FT where (f) 是 (t) 的 continuous Fourier transform G(f) 是 {gk} 的 discrete time Fourier transform

17 連乘 (可以藉由 normalization, 讓 (0) = 1) 若G(f) 決定了,則 (f) 可以被算出來 G(f): 被稱作 generating function constraint 1

18  同理 constraint 2  另外,由於 (f = 0 代入) 必需滿足 constraint 3

19 13-F Real Coefficient Constraints
Since If are satisfied, constraint 4 constraint 5 then (f) = *(− f), (f) = *(−f), and (t), (t) are real. Note: If these constraints are satisfied, gk, hk on page 383 are also real.

20 13-G Vanish Moment Constraint
If (t) has p vanishing moments, for k = 0, 1, 2, …, p−1 Since from the Parseval’s Theorem, (Here, we use the fact that general form of the sifting property )

21 for k = 0, 1, 2, …, p−1 Therefore, Taking the conjugation on both sides, for k = 0, 1, 2, …, p−1 Since if for k = 0, 1, 2, …, p−1 is satisfied, constraint 6 then for k = 0, 1, 2, …, p−1 are satisfied and the wavelet function has p vanishing moments.

22 13-H Orthogonality Constraints
(t): wavelet function If the above equality is satisfied, forward wavelet transform: inverse wavelet transform: (much easier for inverse) C = mean of x(t) (證明於後頁)

23 If and then due to

24 Therefore, is the inverse operation of #

25 ※ 要滿足 之前,需要滿足以下三個條件 (1) for mother wavelet 這個條件若滿足, 對所有的 m 皆成立 (2) for scaling function 嚴格來說,這並不是必要條件,但是可以簡化 第 (3) 個條件的計算

26 (3) for any n, n if k > 0 若 (1) 和 (3) 的條件滿足,則 也將滿足 (Proof): Set If (3) is satisfied, when m  m1 In the case where m = m1, if (1) is satisfied, then #

27  由 Page 396 的條件 (1) Parseval’s theorem if p is an integer Therefore, for all f should be satisfied

28  同理,由 Page 396 的條件 (2) for scaling function 推導過程類似 page 398 for all f should be satisfied

29 衍生的條件:將 代入 (page 398) 因為 hk 是 discrete sequence, H(f) 是hk 的 discrete-time Fourier transform

30 因為 for all f (page 398 的條件) constraint 7

31 同理,將 代入 (page 398) 經過運算可得 constraint 8

32  Page 397 條件 (3) 的處理 由於 是 的 linear combination 是 的 linear combination 是 的 linear combination : 是 的 linear combination 所以 必定可以表示成 的 linear combination

33 所以,若 for any n1, nk 可以滿足 for any n1, nk 必定能夠成立 Page 397 條件 (3) 可改寫成 (將 t − n1 變成 t,  = nk − n1) (from Parseval’s theorem)

34 Since (since from page 404,  is an integer)

35 Since Since for all f (page 398) constraint 9

36 13-I Nine Constraints 整理: 設計 mother wavelet 和 scaling function 的九大條件 (皆由 page 382 的 constraints 衍生而來) (1) for fast algorithm , page 388 (2) for fast algorithm , page 389 (3) for fast algorithm , page 389 (4) for real , page 390 (5) for real , page 389 (6) for p vanish moments , page 392 for k = 0, 1, …, p-1

37 (7) for orthogonal , page 401 for orthogonal , page 402 (8) for orthogonal , page 406 (9)

38  條件的簡化 有時,令 此時,若 (條件 (5), (8) 滿足) 條件 (4), (7), (9) 也將滿足

39 整理: 設計 mother wavelet 和 scaling function 的幾個要求 (簡化版)
(1) for fast algorithm (2) for fast algorithm (3) for fast algorithm (4) for real for p vanish moments (5) for k = 0, 1, …, p-1 (6) for orthogonal (7)

40 13-J Design Process 設計時,只要 G(f) (0  f  1/4) 決定了,mother wavelet 和scaling function 皆可決定 G(f): 被稱作 generating function Design Process (設計流程): (Step 1): 給定 G(f) (0  f  1/4),滿足以下的條件 (a) (b) for k = 0, 1, 2, …, p-1

41 (Step 2) 由 決定G(f) (3/4  f < 1) (Step 3) 由 決定G(f) (1/4 < f < 3/4) 再根據 G(f) = G(f+1),決定所有的 G(f) 值 (Step 4) 由 決定H(f) (Step 5) 由 決定(f), (f)

42 註: (1) 當 Step 1 的兩個條件滿足,由於 for k = 0, 1, 2, …, p-1 又由於 for k = 0, 1, 2, …, p-1 (2) 所以當 G(f) (0  f  1/4) 給定,|G(f)| 有唯一解

43 13-K Several Continuous Wavelets with Discrete Coefficients
(1) Haar Wavelet g[0] = 1, g[1] = 1 h[0] = 1, h[1] = −1 g[0] = 1/2, g[1] = 1/2 h[0] = 1/2, h[1] = −1/2 vanish moment = ?

44 (2) Sinc Wavelet for |f | 1/4 otherwise vanish moment = ? (3) 4-point Daubechies Wavelet vanish moment = ? vanish moment VS the number of coefficients

45 From: S. Qian and D. Chen, Joint Time-Frequency Analysis: Methods and
Applications, Prentice Hall, N.J., 1996.

46 13-L Continuous Wavelet with Discrete Coefficients 優缺點
 Advantages: (1) Fast algorithm for MRA (2) Non-uniform frequency analysis FT (3) Orthogonal

47  Disadvantages: (a) 無限多項連乘 (b) problem of initial 皆由 算出 如何算 (c) 難以保證 compact support (d) 仍然太複雜

48 附錄十三 幾種常見的影像壓縮格式 (1) JPEG: 使用 discrete cosine transform (DCT) 和 88 blocks 是當前最常用的壓縮格式 (副檔名為 *.jpg 的圖檔都是用 JPEG 來壓縮) 可將圖檔資料量壓縮至原來的 1/8 (對灰階影像而言) 或 1/ (對彩色影像而言) (2) JPEG2000: 使用 discrete wavelet transform (DWT) 壓縮率是 JPEG 的 5 倍左右 (3) JPEG-LS: 是一種 lossless compression 壓縮率較低,但是可以完全重建原來的影像 (4) JPEG2000-LS: 是 JPEF2000 的 lossless compression 版本 (5) JBIG: 針對 bi-level image (非黑即白的影像) 設計的壓縮格式

49 (6) GIF: 使用 LZW (Lempel–Ziv–Welch) algorithm (類似字典的建構)
適合卡通圖案和動畫製作,lossless (7) PNG: 使用 LZ77 algorithm (類似字典的建構,並使用 sliding window) lossless (8) JPEG XR (又稱 HD Photo): 使用 Integer DCT,lossless 在 lossy compression 的情形下壓縮率可和 JPEG 2000 差不多 (9) TIFF: 使用標籤,最初是為圖形的印刷和掃描而設計的,lossless


Download ppt "XIII. Continuous WT with Discrete Coefficients"

Similar presentations


Ads by Google