(Back to Session Schedule)

The 31st International Technical Conference on Circuits/Systems, Computers and Communications

Session M3-5  Image Processing (3)
Time: 15:50 - 17:50 Monday, July 11, 2016
Location: Room 5
Chair: Jong-Il Park (Hanyang University, Republic of Korea)

M3-5-1 (Time: 15:50 - 16:10)
TitleAdaptive Mode Selection for Low Complexity Enhancement Layer Encoding of SHVC
Author*Kazuki Kuroda, Takafumi Katayama, Tian Song, Takashi Shimamoto (The University of Tokushima, Japan)
Pagepp. 355 - 358
KeywordSHVC, Merge mode, Motion estimation, Complexity reduction, Coding efficiency
AbstractIn this work, we proposed a low complexity algorithm using the adaptive mode selection based on the sum of squared errors of the correlated pixels and motion vector. To achieve low computation complexity, the temporal correlation and the spatial correlation is used for the mode decision. The proposed algorithm is evaluated by the reference software of SHVC. The simulation results show that the proposed algorithm can achieve over 20% computation complexity reduction comparing to the original SHVC algorithm.

M3-5-2 (Time: 16:10 - 16:30)
TitleProposal of Haze Removal Method with Adjustable Processing Degree
Author*Yi Ru, Go Tanaka (Nagoya City University, Japan)
Pagepp. 359 - 362
Keywordhaze removal, atmospheric scattering, bilateral filter
AbstractImages are often degraded by haze, and the visibility of an image which includes haze is poor. To improve the visibility of such images, some methods have been proposed. In this paper, we propose a new haze removal method which can easily adjust the processing degree by two parameters. It enables avoiding the excessive and inadequate processing problems. In addition, the haze residue in partial area is suppressed in the proposed method. We also use the bilateral filter to sharpen an output image. Experimental results show the effectiveness of the proposed method.

M3-5-3 (Time: 16:30 - 16:50)
TitleExemplar-Based Image Inpainting Using Context-Aware Approach
Author*Yusuke Murakami, Koichi Ichige (Yokohama National University, Japan)
Pagepp. 363 - 366
KeywordExemplar-based image inpainting, Context-aware approach
AbstractThis paper presents a novel exemplar-based inpainting method using context-aware approach. There exist conventional methods which use the positional relations between the most similar patches in patch unit, but it cannot achieve the best performance. By applying context-aware approach to it, we propose an high-quality and efficient exemplar-based inpainting method. Performance of the proposed method is evaluated through computer simulation.

M3-5-4 (Time: 16:50 - 17:10)
TitleAccurate Feature Point Tracking for Omnidirectional Image Sequence
Author*Hiroshi Tada, Koichi Ichige (Yokohama National University, Japan)
Pagepp. 367 - 370
Keywordfeature point tracking, omnidirectional image, optical flow
AbstractIn this paper, we propose a highly accurate feature point tracking method for omnidirectional image sequence. First we try to enlarge the number of optical flows by interpolation. Then we improve the tracking accuracy by combining the flows which will make smaller tracking error. Performance of the proposed method is evaluated through computer simulation.

M3-5-5 (Time: 17:10 - 17:30)
TitleLocal Signal-Dependent Noise Estimation on Texture Domain for CFA Raw Images
Author*Kyu-Ho Lee, Jong-Ok Kim (Korea University, Republic of Korea)
Pagepp. 371 - 374
Keywordimage denoising, signal-dependent noise, local noise estimation, CFA raw image
AbstractIn this paper, we propose a method of signal-dependent noise estimation and denoising that operates on the CFA raw image. The proposed method effectively deals with signal dependent noise by estimating and denoising noise on the texture domain in a CFA LR component-wise and localized manner. The proposed method is practically evaluated on a simulated end-to-end imaging pipeline. The experimental results indicate that the proposed method indeed efficiently removes signal-dependent noise.

M3-5-6 (Time: 17:30 - 17:50)
TitleEmbedded Implementation Model of 2-D Non-separable Oversampled Lapped Transforms for Video Processing
Author*Keita Imai, Shogo Muramatsu (Niigata University, Japan)
Pagepp. 375 - 378
KeywordEmbedded implementation, Non-separable oversampled lapped transform, HW/SW co-implementation, Image processing
AbstractThis work proposes an embedded implementation model of 2-D Non-separable Oversampled Lapped Transform (NSOLT) for video processing. NSOLT has successfully been applied to image restoration. The authors have proposed an embedded implementation model of NSOLT as a previous work. The existing model,however, cannot efficiently process successive frames of a video. To realize efficient video processing, a modified embedded implementation model of NSOLT is proposed by introducing a pipeline architecture.