臺灣博碩士論文加值系統

本論文的目的是提出有效消除雜訊的新方法，以提升恢復影像的品質。目前在影像雜訊消除方面的主流研究方法，可分為兩部份：雜訊檢測與雜訊取代。在雜訊檢測方法中，一個常用的方法是以邊界為基礎的雜訊檢測方法。然而以邊界為基礎的檢測方法若是邊界設定不適當將導致雜訊像素無法消除，以致嚴重影響恢復影像的品質。為了改善這個問題，我們提出了三個以邊界為基礎的雜訊檢測方法：MBDND_1、MBDND_2與NDEND。MBDND_1與MBDND_2主要是修改了文獻[1]的BDND (Boundary Discriminative Noise Detection)雜訊檢測方法中的像素分群不等式，因此我們稱這類方法為MBDND (Modified BDND)。本論文提出的第三種以邊界為基礎的雜訊檢測方法，主要是依照估計雜訊分佈來決定雜訊檢測所需要的邊界。這個方法稱為NDEND (Noise Detection Based on Estimated Noise Distribution)。模擬結果顯示我們所提出的MBDND_1、MBDND_2與NDEND方法的檢測性均能優於BDND。
雜訊取代的主要目的是根據雜訊檢測結果，以非雜訊像素取代雜訊像素。在雜訊取代方面，我們發現以中值濾波為基礎的方法中，使用的視窗大小會影響恢復影像的品質。使用較大的視窗會導致較大的平滑效果，而使用較小的視窗往往會有較好的對比。基於上述的觀察，我們提出適應性近鄰中值濾波器(Adaptive Neighborhood Median Filter, ANMF)的雜訊取代方法。ANMF方法有三種類型分別為ANMF_1、ANMF_k和ANMF_K。第一類型ANMF_1以最近鄰非雜訊像素取代中心雜訊像素，第二類型ANMF_k是以視窗內所有非雜訊像素的中值取代雜訊像素，第三類型ANMF_K則需要視窗內非雜訊像素總數不小於K時，才會以其中值取代中心雜訊像素。模擬結果顯示ANMF_1與ANMF_k 方法具有不錯的恢復影像品質，而ANMF_K方法的結果則取決於K值的大小。此外，在模擬中所提的ANMF方法也與文獻[1]的雜訊取代方法，姑且稱之為MNASM (Modified Noise Adaptive Soft-Switching Median)，進行比較，結果本論文的方法所得到的恢復影像具有較佳的品質與較多的細節。
最後，我們將NDEND檢測方法與ANMF_k 取代方法整合成一個雜訊消除方法，稱為NDEND/ANMF_k 方法。並與文獻[1]的BDND/MNASM雜訊消除方法比較，結果顯示本論文所提的NDEND/ANMF_k 方法，不論在椒鹽或隨機值脈衝雜訊的表現上，均優於BDND/MNASM雜訊消除方法。

This thesis presents a novel impulse noise removal approach to improve the quality of restored image. Presently, noise removal approaches generally consist of two stages: noise detection and noise replacement. In the stage of noise detection, a noisy pixel is identified. If a noisy pixel is detected, a noise replacement scheme is applied to replace the noisy pixel with un-noisy one. When the pixel is uncorrupted, then leave it intact. A well-known noise detection scheme is the boundary discriminative noise detection (BDND). The performance of BDND is heavily dependent on the accuracy of boundary detection. When the boundaries are not determined appropriately, then the noisy pixel will be shown in the restored image. To improve the detection performance of BDND, two modified BDND are proposed in this thesis. They are called MBDND_1 and MBDND_2, respectively. In the MBDND_1, a modification is made on the inequalities of BDND while a boundary resetting scheme is applied in the MBDND_2. Besides, a novel noise detection scheme called the noise detection based on estimated noise distribution (NDEND) is presented and shown having much better detection performance than the BDND, the MBDND_1, and the MBDND_2.
As for the noise replacement, a class of adaptive neighborhood median filters (ANMF) is introduced. Note that the window size used in the filtering process has smoothing effect on the restored image. That is, larger windows applied in the filtering process result in a stronger smoothing effect on the restored image. Thus, the proposed ANMF employs smaller windows, when replacing noisy pixels, for better visual quality of a restored image, especially in high noise density cases. To justify the proposed NDEND and ANFM, several images are given where the salt and pepper noise, the random-valued noise, and the unbalanced density noise, with various densities are under study. Besides, the results obtained from the NDEND/ANFM are compared with the well-known boundary based approach BDND. It indicates that the proposed NDEND/ANFM generally has better performance than that in the BDND both in objective and/or subjective assessments.

中文摘要 I
Abstract III
致謝 V
索引目錄 VI
表目錄 IX
圖目錄 X
第一章 簡介 1
第二章 相關文獻回顧 4
2.1 脈衝雜訊種類說明 5
2.2 雜訊檢測方法回顧 7
2.2.1 模糊理論雜訊檢測方法 7
2.2.2 自組織類神經雜訊檢測方法 8
2.2.3 模糊分類器雜訊檢測方法 8
2.2.4 以Dempster–Shafer理論為基礎的雜訊檢測方法 8
2.2.5 邊界分群雜訊檢測方法 9
2.3 雜訊取代方法回顧 10
2.3.1 使用非雜訊像素之中值濾波 10
2.3.2 適應性軟切換式中值濾波方法 11
2.3.3 改良的適應性軟切換式中值濾波方法 12
2.3.4 方向性加權中值濾波器 13
第三章 以邊界為基礎之雜訊檢測方法 14
3.1 BDND雜訊檢測範例說明 15
3.2 改良式邊界分群檢測方法(MBDND) 17
3.2.1 第一種改良式邊界分群檢測法(MBDND_1) 17
3.2.2 第二種改良式邊界分群檢測方法(MBDND_2) 19
3.3 NDEND雜訊檢測方法 20
3.3.1 NDEND實行步驟 20
3.3.2 NDEND雜訊檢測範例說明 22
3.4 雜訊檢測模擬結果與討論 24
3.4.1 椒鹽雜訊 24
3.4.2 隨機值雜訊 28
3.4.3 不平衡椒鹽雜訊 34
第四章 適應性雜訊取代方法 39
4.1 MNASM雜訊取代範例 39
4.2適應性近鄰中值濾波器 41
4.2.1 ANMF_1雜訊取代方法 42
4.2.2 ANMF_k雜訊取代方法 43
4.2.3 ANMF_K雜訊取代方法 43
4.2.4 ANMF雜訊取代範例 44
4.3模擬結果與討論 45
4.3.1 客觀PSNR之比較 46
4.3.2 主觀視覺之比較 47
4.4 雜訊消除模擬結果與討論 51
4.4.1 椒鹽雜訊 52
4.4.2 隨機值雜訊 53
4.4.3 不平衡椒鹽雜訊 57
第五章 結論 60
參考文獻 62


表目錄
表3.3.1 h V 直方圖統計表.............................................................................. 23
表3.4.1 MBDND 與NDEND 跟BDND 之MD 與FA 比較結果(Boat) ....... 26
表3.4.2 MBDND 與NDEND 跟BDND 之MD 與FA 比較結果(Goldhill).. 27
表3.4.3 MBDND 與NDEND 跟BDND 之MD 與FA 比較結果(Barbara) .. 28
表3.4.4 MBDND_1 與MBDND_2 之MD 與FA 比較結果(Boat)................ 29
表3.4.5 MBDND_1 與MBDND_2 之MD 與FA 比較結果(Goldhill). ......... 30
表3.4.6 MBDND_1 與MBDND_2 之MD 與FA 比較結果(Barbara)........... 31
表3.4.7 NDEND 與BDND 之MD 與FA 比較結果(Boat)............................ 32
表3.4.8 NDEND 與BDND 之MD 與FA 比較結果(Goldhill) ...................... 33
表3.4.9 NDEND 與BDND 之MD 與FA 比較結果(Barbara)....................... 34
表3.4.10 不平衡雜訊下MBDND 與NDEND 跟BDND 之MD 與FA 比較
(Boat) ............................................................................................................... 36
表3.4.11 不平衡雜訊下MBDND 與NDEND 跟BDND 之MD 與FA 比較
(Goldhill) ......................................................................................................... 37
表3.4.12 不平衡雜訊下MBDND 與NDEND 跟BDND 之MD 與FA 比較
(Barbara) .......................................................................................................... 38
表4.4.1 圖(4.4.3)的相對應隨機值雜訊範圍表. ............................................ 55
表4.4.2 圖(4.4.5)的相對應不平衡雜訊範圍表............................................. 58


圖目錄
圖2.1.1 摻入5%椒鹽雜訊的Lena 影像與直方圖.......................................... 6
圖2.1.2 摻入40%隨機值雜訊的Lena 影像與直方圖.................................... 7
圖2.3.1 適應性切換式雜訊取代結構圖........................................................ 12
圖4.3.1 ANMF 與MNASM 雜訊取代客觀PSNR 之比較圖........................ 47
圖4.3.2 影像Boat 與摻雜雜訊之不同方法所得到的恢復影像.................... 49
圖4.3.3 影像Goldhill 與摻雜雜訊之不同方法所得到的恢復影像............... 50
圖4.3.4 影像Barbara 與摻雜雜訊之不同方法所得到的恢復影像............... 51
圖4.4.1 NDEND/ANMF_k 與 BDND/MNASM 恢復影像PSNR 比較圖.. 53
圖4.4.2 Barbara 恢復影像結果圖(80% 椒鹽雜訊) ...................................... 53
圖4.4.3 NDEND/ANMF_k 與 BDND/MNASM 恢復影像PSNR 比較圖.. 56
圖4.4.4 Barbara 恢復影像結果圖(Case 2)...................................................... 56
圖4.4.5 NDEND/ANMF_k 與 BDND/MNASM 恢復影像PSNR 比較圖.. 59
圖4.4.6 Barbara 恢復影像結果圖(80%胡椒雜訊+10%鹽雜訊) ................... 59

[1]P.-E. Ng and K.-K. Ma, “A Switching Median Filter With Boundary Discriminative Noise Detection for Extremely Corrupted Images,” IEEE Transactions on Image Processing, Vol. 15, No. 6, pp. 1506-1516, June 2006.
[2]I. Pitas and A. N. Venetsanopoulos, “Order Statistics in Digital Image Processing,” Proceedings of the IEEE, Vol. 80, No. 12, pp. 1893-1921, December 1992.
[3]D. R. K. Brownrigg, “The Weighted Median Filter,” Commun. ACM, Vol. 27, No. 8, pp. 807-818, Aguest 1984.
[4]S.-J. Ko and Y. H. Lee, “Center Weighted Median Filters and Their Applications to Image Enhancement,” IEEE Transactions on Circuits and Systems, Vol. 38, No. 9, pp. 984-993, September 1991.
[5]H. Hwang and R. A. Haddad, “Adaptive Median Filters: New Algorithms and Results,” IEEE Transactions on Image Processing, Vol. 4, No. 4, pp. 499-502, April 1995.
[6]T. Chen, K.-K. Ma, and L.-H. Chen, “Tri-State Median Filter for Image Denoising,” IEEE Transactions on Image Processing, Vol. 8, No. 12, pp. 1834–1838, December 1999.
[7]Z. Wang and D. Zhang, “Progressive Switching Median Filter for the Removal of Impulse Noise from Highly Corrupted Images,” IEEE Transactions on Circuits and Systems, Vol. 46, No. 1, pp. 78–80, January 1999.
[8]S. Zhang and M. A. Karim, “A New Impulse Detector for Witching Median Filters,” IEEE Signal Processing Letters, Vol. 9, No. 4, pp. 360-363, November 2002.
[9]G. Pok, J.-C. Liu, and A. S. Nair, “Selective Removal of Impulse Noise Based on Homogeneity Level Information,” IEEE Transactions on Image Processing, Vol. 12, No. 1, pp. 85-92, January 2003.
[10]Wei Ping, Li Junli, Lu Dongming and Chen Gang, “A Fast and Reliable Switching Median Filter for Highly Corrupted Images by Impulse Noise,” IEEE International Symposium on Circuits and Systems, pp. 3427-3430 May 2007.
[11]R. H. Chan, C.-W. Ho, and M. Nikolova, “Salt-and-Pepper Noise Removal by Median-Type Noise Detectors and Detail-Preserving Regularization,” IEEE Transactions on Image Processing, Vol. 14, No. 10, pp. 1479-1485, October 2005.
[12]S. Schulte, M. Nachtegael, V. D. Witte, D. V. D. Weken, and E. E. Kerre, “A Fuzzy Impulse Noise Detection and Reduction Method,” IEEE Transactions on Image Processing, Vol. 15, No. 5, pp. 1153-1162, May 2006.
[13]H. Kong and L. Guan, “A Noise-Exclusive Adaptive Filtering Framework for Removing Impulse Noise in Digital Images,” IEEE Transactions on Circuits and Systems, Vol. 45, No. 3, pp. 422-428, March 1998.
[14]Jyh-Yeong Chang and Jia-Lin Chen, “Classifier-Augmented Median Filters for Image Restoration,” IEEE Transactions on Instrumentation and Measurement, Vol. 53, No. 2, pp. 351-356, April 2004.
[15]Tzu-Chao Lin and Pao-Ta Yu, “Thresholding Noise-Free Ordered Mean Filter Based on Dempster-Shafer Theory for Image Restoration,” IEEE Transactions on Circuits and Systems, Vol. 53, No. 5, pp. 1057-1064, May 2006.
[16]Haosong Kong and Ling Guan,“An Adaptive Approach for Removing Impulsive Noise in Digital Images,” IEEE International Conference on Acoustics, Speech, and Signal Processing, Atlanta, GA, Vol. 4, pp. 2289-2292, 1996.
[17]H.-L. Eng and K.-K. Ma, “Noise Adaptive Soft-Switching Median Filter,” IEEE Transactions on Image Processing, Vol. 10, No. 2, pp. 242-251, February 2001.
[18]Yiqiu Dong and Shufang Xu, “A New Directional Weighted Median Filter for Removal of Random-Valued Impulse Noise,” IEEE Transactions on Image Processing Letters, Vol. 14, No. 3, pp. 193-196, March 2007.