基於補丁之視訊超解析度處理-使用二維隱藏馬克夫模型

Patch-Based Video Super-Resolution Using 2-D Hidden Markov Model

全柏翰(Po-Han Chuan)

指導教授 ： 謝禎冏

大同大學/電機資訊學院/資訊工程學系所/碩士(2010年)

摘要

隨著科技的進步，人們對於影像品質的要求也越來越高，市場上高解析顯示器的價格逐漸平易近人。雖然高解析顯示裝置已經日漸普及，但視訊的品質卻受限於頻寬、播放器、訊號源等，很難在高解析顯示裝置上有令人滿意的品質。高解析度的影像有較密的顯示畫素，所能表示的資訊更加細微，為了能有效的顯示更細微的影像資訊，利用影像處理技術來提升解析度早已開始研究。傳統的影像放大處理所面臨的最大問題就是影像在放大後，會使得影像中物體邊緣有鋸齒狀、模糊等失真問題，因而造成解析度下降，精確度不足。基於以上因素，超解析度法因而提出，目的在於提升影像的解析度。本論文提出一種基於補丁的超解析度演算法，使用二維隱藏馬可夫模型來建立出最合適的超解析影像，首先學習出高解析補丁與低解析補丁之間的相對應關係，並儲存至資料庫中，從中可以得到二維隱藏馬可夫模型中的初始機率π和觀察機率B，當輸入的低解析影像經過初始放大並分割為補丁後，在低解析補丁資料庫中先尋找前五名最為相似的補丁，使用傳統的維特比演算法(Viterbi algorithm)找出最佳相對應的高解析補丁，其間可使用低解析補丁的重疊部分來做相容度的檢驗，算出轉換機率A，以找出最合適的高解析補丁。我們進行了多組實驗包含室內、室外、書本文字和月曆文字，在實際觀察下可發現影像的鋸齒和模糊現象有所改善，在客觀的量化訊號雜訊比(PSNR)，本論文的方法平均約26.26%，比起最高優先貼回得到的平均數值約25.07%，以及Bi-Cubic內插法得到的平均數值約24.82%，有明顯的改善，證實我們提出方法的可行性。

關鍵字

超解析 ； 補丁 ； 二維隱藏馬克夫模型 ； 維特比演算法

並列摘要

With the progress of display technology, consumer tends to demand for higher image quality. As the price of high-resolution display goes down gradually, LCD-TV becomes very popular in the market. However, users may not be satisfied by viewing low resolution images on high-resolution device due to the limited information in the low-resolution source video signal. Traditional image enlargement methods would produce zigzags and blurrings effects. Therefore it will decrease the accuracy of image. Super-resolution technologies were developed to enhance the resolution of low resolution videos. This paper presents a patch based super-resolution algorithm which learns the correspondence of high frequency information between low resolution patches and high resolution patches. The 2-D Hidden Markov Model (HMM) is deployed to find the optimal super resolution image. Firstly, the learned patches are saved into the patch database. Then, the low-resolution video is input for up scaling. For each input patch, search the top five similar patches in the low-resolution database The corresponding high resolution patches are then used to calculate the π and B probabilities for the 2D-HMM. The transition probabilities among the guessed high resolution patches can be verified by the overlapped regions of the corresponding low resolution patches. Finally, Viterbi algorithm is applied to find the best high resolution patches. Several experiments including indoor and outdoor are tested. The PSNR is 26.26% in average which is higher than the other super resolution methods. The results demonstrated the feasibility of proposed method.

並列關鍵字

Super-resolution ； Patch ； 2-D Hidden Markov Model ； Viterbi Algorithm

參考文獻

[2] R. G. Keys, “Cubic Convolution Interpolation for Digital Image Processing,” IEEE Trans. Acoustics, Speech, and Signal Processing, vol. ASSP-26, no. 6, pp. 1153-1160, Dec.1981.

[4] W. K. Pratt, Digital Image Processing, 2nd ed. New York: Wiley, 1991.

[5] M. Unser, A. Aldroubi and M. Eden, “B-Spline Signal Processing: Part II-Efficient Design and Applications,” IEEE Trans. Signal Processing, vol. 41, pp. 834-848, Feb.1993.

[6] M. Unser, A. Aldroubi and M. Eden, “Enlargement or Reduction of Digital Images with Minimum Loss of Information,” IEEE Trans. on Image Processing, vol. 4, pp. 247-258, Mar. 1995

[7] M. Lehmann, Claudia Gonner and Klaus Spitzer, “Addendum: B-spline Interpolation in Medical Image Processing,” IEEE Trans. on Medical Image, vol. 20, pp. 660-665, July 2001.