論文 - 椋木 雅之
-
A unified camera calibration using geometry and blur of feature points 査読あり
M. Baba; M. Mukunoki; N. Asada
Proceedings - International Conference on Pattern Recognition 1 816 - 819 2006年
記述言語：日本語 掲載種別：研究論文（国際会議プロシーディングス）
-
A new camera calibration method taking blur effects into account 査読あり
Baba M., Mukunoki M., Asada N.
ACM SIGGRAPH 2005 Sketches, SIGGRAPH 2005 2005年7月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：ACM SIGGRAPH 2005 Sketches, SIGGRAPH 2005
-
Reflectance estimation of sparkle in metallic paints 査読あり
Baba M., Miura S., Mukunoki M., Asada N.
ACM SIGGRAPH 2005 Posters, SIGGRAPH 2005 2005年7月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：ACM SIGGRAPH 2005 Posters, SIGGRAPH 2005
-
3D model generation from image sequences using global geometric constraint 査読あり
M. Mukunoki; K. Yasuda; N. Asada
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 3804 LNCS 470 - 477 2005年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス）
DOI： 10.1007/11595755_57
-
Retrieval of images using pixel based object models 査読あり
Mukunoki M., Minoh M., Ikeda K.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 945 557 - 566 2005年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
© Springer Verlag 2005. We apply our Pixel-based object labeling method tothe problem of indexing images. Our method is a means to assign an object label to each pixel in out-door scenes. It is suitable for automatic object labeling and applicable to automatic indexing problems. The recognition rate of our method is about 77%. To apply our method to the problem of indexing, we introduce a pictorial query method for retrieva.We conducted some retrieval experiments, and examined the usability of our method to the indexing problem.
-
Large object shape recovery from uncalibrated camera motion by non-linear optimization 査読あり
Asada N., Mukunoki M., Migita T., Aoyama M.
Sixth IASTED International Conference on Signal and Image Processing 151 - 156 2004年12月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Sixth IASTED International Conference on Signal and Image Processing
3D recovery of object shape and camera motion from 2D image sequence is formulated as a non-linear optimization problem. Such a non-linear approach ensures the most precise solution, but it involves computational difficulties, i.e. complexity and instability. This paper describes a practical procedure to solve the non-linear optimization problem for object shape and camera motion recovery. First, we discuss the imaging model for large objects such as buildings; one shot and pseudo wide angle models, where the camera motion is modeled as a uniform motion along circular path with gazing at the center. Then, the generalized initial values for circular motion is proposed and "mean appearance ratio (MAR)" of feature points is defined as a measure of stability. Experimental results with synthetic images show that 3D recovery was stably performed from image sequences with MAR over 15%, quasi-stable with over 8.5% for one shot model, and stable with over 10.3%, quasi-stable with over 4. 4% for pseudo wide angle model. 3D shape and motion using real images were recovered successfully from MAR 17% for one shot model and 6% for pseudo wide angle one. These results have demonstrated the robustness of the generalized initial values and effectiveness of MAR for large object shape from uncalibrated camera motion problem.
-
Modification table form generation system based on the form recognition 査読あり
Amano A., Asada N., Mukunoki M.
Proceedings - International Conference on Pattern Recognition 2 659 - 662 2004年12月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Proceedings - International Conference on Pattern Recognition
As there exist large number of both printed and electronical documents of table form, it is very important to provide their handling system. The system should provide capability of generating, modifying or filling in to the form only with logical manipulation. In this paper, we propose a system which can extract the structure of the form, and modify it logically without considering layout information, and finally generate modified form layout. The system is based on our table form analysis method and table form representation language (TFML). Experimental results show that the system can generate practical layouts for the modified documents.
-
Estimating roughness parameters of an object's surface from real images 査読あり
Baba M., Mukunoki M., Asada N.
ACM SIGGRAPH 2004 Posters, SIGGRAPH 2004 2004年8月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：ACM SIGGRAPH 2004 Posters, SIGGRAPH 2004
-
Shadow removal from a real image based on shadow density 査読あり
Baba M., Mukunoki M., Asada N.
ACM SIGGRAPH 2004 Posters, SIGGRAPH 2004 2004年8月
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：ACM SIGGRAPH 2004 Posters, SIGGRAPH 2004
-
“Shadow Removal from a Real Image Based on Color Clustering” 査読あり
Baba M., Mukunoki M., Asada N.
Journal of the Institute of Image Electronics Engineers of Japan 33 ( 3 ) 531 - 538 2004年1月
記述言語：日本語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Journal of the Institute of Image Electronics Engineers of Japan
This paper proposes a method to remove shadows from a real image based on RGB color space analysis. The algorithm consists of five steps. First, all pixel colors are mapped on the RGB space, and the color distribution is segmented by using clustering method. Second, a pair of clusters that are almost aligned on a color vector is found. We consider the darker cluster as shadow area and the lighter as sunshine one. Next, we perform the color adjustment between them by modifying the direction of shadow cluster to be that of sunshine one. Fourth, the color average and variance of the shadow cluster are adjusted to be the same as those of the sunshine one. Finally, we apply a smoothing filter to correct the boundary discontinuity that sometimes occurs around fuzzy boundary between shadow and sunshine areas. Experimental results show that our algorithm worked successfully inreal situations. © 2004, The Institute of Image Electronics Engineers of Japan. All rights reserved.
-
Controlling a Camera with Minimized Camera Motion Changes under the Constraint of a Planned Camera-work 査読あり
Yasutaka ATARASHI,Yoshinari KAMEDA,Masayuki MUKUNOKI,Koh KAKUSHO,Michihiko MINOH,Katsuo IKEDA
International Workshop on Pattern Recognition and Understanding for Visual Information Media in Cooperation with ACCV 2002 2002年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス）
-
Classification of rawmaterial sports videos for broadcasting using color and edge features 査読あり
Mukunoki M., Bettini M., Assfalg J., Del Bimbo A.
Proceedings - IEEE International Conference on Multimedia and Expo 1187 - 1190 2001年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Proceedings - IEEE International Conference on Multimedia and Expo
We discuss the method to classify raw material sports videos for broadcasting. Because the raw material sports videos sometimes do not edited, we cannot use the knowledge on edited videos. We use the color and edge features and evaluate whether we can classify the sports videos with those features. We also introduce "player" and "audience" class apart from each sport class to improve the classification results.
-
Detection of the indicated area with an indication stick 査読あり
Sakiyama T., Mukunoki M., Ikeda K.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 1948 480 - 487 2000年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
© Springer-Verlag Berlin Heidelberg 2000. This paper describes a detection method of indication action and the corresponding indicated area with a stick. In human communication, there exists ambiguity of indication action, so we need some useful knowledge about indication action. We propose a detection method of the indicated area based on the knowledge which we clarify by means of observation of some lectures. We formulate a potential value of intentional indication action of an instant, and cast the value for printed area using a weighted vote method. An experimental result shows that 85 percents of indication actions are correctly detected.
-
Classification method of images based on composition and its application to image retrieval 査読あり
Maeda S., Mukunoki M., Ikeda K.
International Conference on Multimedia Computing and Systems -Proceedings 2 240 - 244 1999年
記述言語：英語 掲載種別：研究論文（国際会議プロシーディングス） 出版者・発行元：International Conference on Multimedia Computing and Systems -Proceedings
We propose a visual query method for image retrieval, in which the user expresses the composition of the target image by selecting one of the composition types presented by the system, and also propose an image classification method to derive composition types from an image database. From the viewpoint of communication between the system and its user, we point out the problem of visual query methods using concrete images as queries. Though suited to convey visual image properties, they grow the ambiguity in the system's interpretation of queries. To manage this tradeoff, we implemented our methods in a prototype system and derived composition types from an image database to show how our query method works in image retrieval.