近似動向補償之彩色視訊脈衝雜訊移除

為了兼顧雜訊濾除及細節保存，切換式影像雜訊抑除方法廣受學界青睞。而視訊去雜訊時更經常會加入前後像幀關聯，甚至導入動向預測補償。這些組合方法雖然具有卓越的復原表現，卻經常因為過於繁雜的轉換或大範圍的搜尋而耗時太久，無法滿足快速的視訊實務應用。本文提出一新型的近似動向補償去雜訊演算法(PMCDA)，期望在抑制視訊雜訊過程中取得較好的性能與效率折衷。我們首先衍伸簡易的同類群集偵測法為3D版本，藉以排除明確的非雜訊像素不處理來加速預偵測程序。因為提出的三維同類群集偵測法不需任何的排序或動向補償，在較常見的低雜訊情況，能大幅降低去雜訊方法整體的計算成本。PMCDA同時最大化預偵測之效益，在高雜訊時將其餘候選雜訊像素依其立體周圍可參考程度排序進行第二階段濾波。優先處理較可信區域不僅可以獲得較好還原，亦可以單次濾波成本模擬多次濾波效果，有助於雜訊較高區域獲得可靠參考，同時改善邊框破壞。我們並可以根據預偵測結果評估視訊雜訊率，選用適當之第二階段濾波方法。特別是我們同時提出一新型、簡易且快速的近似動向補償架構，讓新演算法確實取得更好的性能與效率折衷。實驗結果證明，新方法在維持快速優勢的同時，仍能表現出較多數經典視訊去雜訊方法優異的效能。

關鍵字

視訊去雜訊；可參考程度；脈衝雜訊；動向補償

並列摘要

In order to take into account the both objects of noise suppression and detail preservation, the switching image denoising method is widely used by academics. Moreover, the strong correlation between adjacent frames, especially the accurate motion-prediction compensation, is also introduced into the video denoising scheme. While these compounded methods produce good performance, they are either unsuited for hardware implementation due to complicated transformation or time consuming due to large-range search. These defects limit the usage of new results in rapid practical applications of video. This thesis proposed a novel Pseudo-Motion-Compensated Denoising Algorithm (PMCDA) to achieve a better trade-off between the performance and the efficiency during the noise suppression process of the video. First, the simple peer-group detection method is extended to a 3D version to accelerate the pre-detection process and ignore the precisely noise-free pixels. It’s noted that neither any sort nor move compensation is needed in the 3D peer-group detection. Therefore, the fast pre-detection stage can significantly reduce the overall cost of computing since low-noise contamination is more common in real-life videos. In the high-noise circumstance, PMCDA can maximize the utilization of the detection. The remaining noise candidates will be filtered via a sequence according to their 3D noise ratio. The more reliable area has higher priority in process to get better restoration. It can also achieve a similar effect as that of recursive method by only using a cost of single round and improve the border damage by providing more reliable references to high-noise area. Furthermore, based on the pre-detection results, we can also evaluate the rough noise ratio of video and to decide the appropriate filter for the second stage. In particular, we propose a new, simple and fast pseudo-motion-compensation methodology. As a result, the new method does achieve better performance and efficiency tradeoffs. The experimental simulations show that the proposed video denoising algorithm still outperforms some other state-of-the-art methods while exhibiting the advantages of fast computation.

並列關鍵字

Video Denoing ； Referable level ； Impulse Noise ； Motion Compensation

參考文獻

[1] A. Buades, B. Coll, and J. M. Morel, “A Review of Image Denoising Algorithms, with a New One”, Multiscale Model. Simul., vol.4 no.2, pp. 490–530, 2005.

[3] A. Buades, J.-L. Lisani, and M. Miladinovi´, “Patch-Based Video Denoising With Optical Flow Estimation”, IEEE Trans. Image Process., vol. 25, no. 6, pp. 2573-2586, 2016.

[4] H. Li, C.Y. Suen, “A novel Non-local means image denoising method based on grey theory”, Pattern Recognition, vol. 49, pp. 237-248, 2016.

[5] L. Bao, Q. Yang, and H. Jin, “Fast Edge-Preserving Patch Match for Large Displacement Optical Flow”, IEEE Trans. Image Process., vol. 23, no. 12, pp. 4996-5006, 2014.

[7] Y. Dong and S. Xu, “A New Directional Weighted Median Filter for Removal of Random-Valued Impulse Noise” IEEE Signal Process. Lett., vol. 14, no. 3, pp. 193–196, 2007.

國際替代計量

近似動向補償之彩色視訊脈衝雜訊移除

全文下載

主題瀏覽