The demand for visual entertainment and high-quality images is increasing with continuous technological advancement and popularity of digital imaging products. Most digital cameras available in the market carry anti-shaking and auto-focus functions, yet even professional photographers may face the problem of blurred images under insufficient lighting. Image processing software currently available in the market usually rectifies blurriness by just sharpening the edges of images, which is not the best solution. Many companies worldwide are working hard on developing deblurring technology, yet there is still a long way to go until application software can be developed. The estimation method used in the academic arena will even take several hours and much adjustment on data is needed to achieve image deblurring.

 

Prof. JIA Jiaya of the Department of Computer Science and Engineering at The Chinese University of Hong Kong (CUHK) has developed a world's leading image deblurring technology to rectify blurriness in photos caused by lighting and camera shaking. Compared to the image sharpening technology, this new technique is more effective and efficient by increasing the image clearness by 100 times, and not more than 10 seconds are needed to process a photo with 1,000,000 pixels, which is a real breakthrough in image restoration technology.

 

Professor Jia and his post-doctoral fellow Dr. Xu Li proposed a novel kernel estimation method based on the integrated study on speed, reliability and user-friendliness, and took the noise control and ringing artifacts into consideration for restoring pictures from significant motion blur. The research team is able to efficiently and accurately reconstruct the blurry data in the recorded image with adoption of deconvolution, analysis of blur kennel and identification of the camera motion.

 

The research started last year and patent has been obtained in the US and in mainland China. Professor Jia's team ran continuous tests on various kinds of images, including night shots and shots of sports, wild animals and aquarium with positive results. Professor Jia is now looking forward to wider daily application of the technology and he also hopes it will help restore and preserve important historical visual data.

 

Please visit the following website for examples of deblurred outputs: www.cse.cuhk.edu.hk/~leojia/projects/robust_deblur.

 

隨著科技不斷進步，數碼產品日益普及，大眾對視覺享受及圖像質素的要求愈來愈 高。現時的數碼相機大多具備自動防震及影像對焦功能。然而，當環境光源不佳 時，專業級攝影師亦有機會因手部搖動而拍出模糊的照片。綜觀目前市面上的應用 圖像軟件，一般只會用圖像邊緣銳利化來處理模糊的照片，這絕非真正還原圖像之 上策。世界各大應用軟件公司正在積極鑽研圖像去模糊技術，然而有關技術仍未成 熟至可轉化為實際應用；學術領域的去模糊算法則動輒需要數十分鐘至數小時進行 演算，並需調制大量參數。

 

香港中文大學（中大）計算機科學與工程學系賈佳亞教授及其指導的博士後研究生 徐立研發領先全球的圖像去模糊技術。相比於圖像邊緣銳利化的功能，這項新技術 的去模糊效果更為顯著及省時，提升圖像清晰度高達100倍，處理100萬像素圖像 的速度少於10秒，是全球圖像處理的一大重要突破。

 

賈教授及其研究團隊提出一項去除圖像震動性模糊的演算法，特別針對速度、可靠 性和易用性等多方面進行綜合研究。研究團隊分析抑制雜訊及震盪效應，從照片中 選取資訊含量最高的結構作出估測，並計算相機震動的軌跡及模糊核心，重建影像 數據，令模糊的照片迅速回復清晰，大幅提升圖像的整體品質。

 

是項研究由去年開始進行，有關技術已成功於美國及國內註冊專利。研究人員利用 此圖像處理程式測試大量不同種類的照片，包括夜景、體育運動、野生動物及水族 館等照片，證實此技術對修復各種類型的模糊圖像皆有顯著功效。賈教授期望將此 軟件推出巿場，讓用戶能夠輕易處理及修復日常的模糊照片。此外，他亦希望藉此 修復別具歷史價值的影像。

 

有關圖像去模糊軟件的成功例子，請瀏覽以下網站： www.cse.cuhk.edu.hk/~leojia/projects/robust_deblur.