A First Parallel Multi-Direction Switching Median Filter for Reducing Impulse Noises

2015 Fiscal Year Final Research Report

• Project/Area Number: 25330218
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Perceptual information processing
• Research Institution: Nagano National College of Technology
• Principal Investigator: Miyazaki Takashi 長野工業高等専門学校, その他部局等, 教授 (10141889)
• Co-Investigator(Renkei-kenkyūsha): YAMAMOTO Hiroaki 信州大学, 工学部情報工学科, 教授 (10182643) ; SONE Mitsuo 東海大学, 情報理工学部, 教授 (40187894)
• Research Collaborator: 横山 靖樹 長野工業高等専門学校, 技術支援部, 技術職員
• Project Period (FY): 2013-04-01 – 2016-03-31
• Keywords: スイッチングメディアンフィルタ / インパルス性雑音 / しきい値 / 並列処理 / GPU / 復元器 / 検出器

Outline of Final Research Achievements

Many kinds of switching-type denoising filters have been proposed to detect and denoise noisy pixels in an image effectively. Our multi-directional switching median filter uses a 2x2 pixels noise detector and an averaging method. To discriminate noisy pixels in the image, our filter requires a threshold like other switching-type filters. The optimal value of the threshold varies depending on the strength and quantity of the edge of the image, which means the optimal threshold varies from area to area of the same image. Focusing on this point, we proposed a new method. It divides the image into segments and, for each of them, sets a threshold proportional to the average edge strength of the segment. Furthermore, we developed the method of the parallel computation using GPU. Our experiment verified that this method could improve image quality.

Free Research Field

画像処理