馬賽克影像之除雜訊及壓縮

Noise Reduction and Compression for Mosaic Image

胡威仲(Wei-Chung Hu)

指導教授 ： 貝蘇章

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.02706

摘要

現今的數位相機和智慧型手機等裝置中，多使用單一感光元件覆蓋一層色彩濾波器陣列的方法來作影像的擷取。為了計算在這種架構下產生的馬賽克圖片所丟失的像素值，就發展出了許多去馬賽克演算法。而本論文則是針對馬賽克圖片和去馬賽克演算法所產生出的兩個議題：ㄧ、去馬賽克前去雜訊。二、壓縮馬賽克圖片。
2006年由 Hirakwa 和 Parks 提出針對馬賽克圖片作去雜訊的想法後，在2009年 Zhang et al. 提出考慮馬賽克圖片特性，在做去馬賽克前先做去雜訊的動作，可以得到較去馬賽克後再去雜訊的程序更佳的影像品質。這主要肇因於若先執行去馬賽克，所得全彩圖片中會包含許多難以去除的鬼影。而本篇論文則提出一種針對色彩濾波器陣列排列方式，進一步所得的改良型非局部平均雜訊消除演算法，並搭配高斯函數分解馬賽克圖片的方式，使所得的全彩圖片影像品質更佳。
對於壓縮馬賽克圖片的部分，2011年 Chung et al. 提出一針對RGB至YUV色域變換矩陣特性及前端色彩濾波器陣列排列方式之關係的修正型色度分量子取樣策略，這樣的方式在解壓縮後的全彩圖片影像品質比傳統的子取樣方式更好。而本篇論文則是在做色域變換前，以頻率域去馬賽克的方式取代時間域去馬賽克法，這種新的壓縮程序會使得解壓縮後的圖片在影像品質和PSNR上都有所提升。

關鍵字

去馬賽克演算法 ； 色彩濾波器陣列 ； 非局部平均雜訊消除演算法 ； 色域變換 ； 色度分量子取樣

並列摘要

Today most of digital cameras and smartphones capture images by single sensor covered with Color Filter Array (CFA). In order to estimate the lost pixel values of mosaic images during the sampling process, many Demosaicking Algorithms were developed. In this paper, we focus on two important issues extended from mosaic image and demosaicking algorithm: 1. Denoising before demosaicking. 2. Compressing mosaic image.
In 2006, Hirakawa and Parks presented the idea of mosaic image noise reduction. In 2009 Zhang et al. proposed a method, instead of denoising after demosaicking, they applied denoising before demosaicking scheme and the result has better visual quality. This is because if we apply demosaicking before denoising, the noisy sensor readings are roots of many color artifacts in demosaicked images and those artifects are hard to remove. In this paper, we proposed a modified non-local means denoising algorithm based on the CFA structure, and also applied Gaussian function to decompose mosaic image. Our method has better visual quality in demosaicked full color image.
For compressing mosaic image, Chung et al. proposed a modified chroma subsampling strategy based on RGB to YUV color domain transform and the structure of CFA. This kind of strategy will give a better visual quality than conventional method. In this paper, we replace spatial demosaicking method with frequency demosaicking before color domain transform. The new compression scheme that we proposed will provide higher PSNR value and good visual quality.

並列關鍵字

Demosaicking Algorithm ； Color filter Array ； Non-local Means Denoising ； Color Domain Transform ； Chroma Subsampling

參考文獻

[2] R. Lukac and K. N. Plataniotis, “Color filter arrays: Design and performance analysis,” IEEE Trans. Consumer Electronics, vol. 51, no. 4, pp. 1260-1267, 2005.

[3] K. Jensen and D. Anastassiou, “Subpixel edge localization and interpolation of still image,” IEEE Trans. On Image Processing, Vol. 4, pp. 285-295.

[5] B. K. Gunturk, Y. Altunbasak and R. M. Mersereau, “ Color plane interpolation using alternating projection,” IEEE Transactions on Image Processing, vol. 11, no.9, 2002.

[6] J. E. Adams and J. F. Hamilton, “Design of practical color filter array interpolation algorithms for digital cameras,” Proc. SPIE, vol. 3028, pp. 117-125, 1997.

[7] E. Dubois, “Frequency-domain methods for demosaicking of Bayer-sampled color images,” IEEE Signal Processing Letters, vol. 12, no. 12, pp. 847-850, 2005.