DEIM Forum 2011 E9-4 252-0882 5322 252-0882 5322 E-mail: t09651yt, sashiori, kiyoki @sfc.keio.ac.jp CBIR A Meaning Recognition System for Sign-Logo by Color-Shape-Based Similarity Computations for Images YUKI TOYOSHIMA SHIORI SASAKI and YASUSHI KIYOKI Faculty of Enviroment and Information Studies, Keio University Graduate School of Media and Governance, Keio University E-mail: t09651yt, sashiori, kiyoki @sfc.keio.ac.jp 5322 Endo,Fujisawa,Kanagawa,252-0882 Japan 5322 Endo,Fujisawa,Kanagawa,252-0882 Japan Abstract In this paper, we present a meaning recognition system for sign-logos in real space. First, this system recognizes the category of input sign-logo images by the similarity computations with images in the database focusing on the color and shape features of images. Second, the system searches for the information corresponding to the specific sign-logo images. By using this system, a user is able to find out the meaning and related information of sign-logos based on the user s location as well as the information of sign-logos around the world. Keyword multimedia database, image database, similarity computation, CBIR 1.
2. [3] / [4] 2 [4] 3. 3.1 1 JIS [3] / 3 (1) (CBIR) (2) CBIR GPS
Input image (1) Image retrieval by color- shape- based similarity computation (CBIR) Provision of Candidate images (2) Image related information selection Display of Sign- logo- related information &meanings 3 Image DB Metadata DB 4 1 (1)CBIR 2 3 4 5 Step1-1) Image processing & Metadata extraction for color (pre- processing) Step1-2) Image processing & Metadata extraction for shape (preprocessing) Step2) Similarity computing using color and shape Step3) Providing candidates by ranking image Step4) Image selection by user Step5) Providing information for selected image Pillar Algorithm K-Means K-means [5][6] 4. (1) (CBIR) (2) (1) 2.2 (CBIR) [5][6] 2 Step 1-1, Step 1-2 Step 2, Step 3 (2) 1 PostgreSQL 2 Step 5 (image ID) (type) (category) (country) (meaning) 5 1 1 4 3.2 Pillar Algorithm [5][6] [5][6] Pillar Algorithm
6 5. 5 227 176 9 10 1 30 public, road, product 14 5.1 5 CBIR 5 (1)CBIR (weight) Color Weight Shape Weight Structure Weight (2) 1 PostgreSQL 6 (1) (2) (3) 10 227 7
1 9 2 7 1 227 10 86.17 46.52 38.55 2 7 (1) 86.17 46.52 38.55 (1) (2) (3) (1) (2) (1) (3) 1 10 5.2 2 2 2 3 4 3 4 similarity [5][6] 2 3 4 3 nosmoking.jpg 4 nosmoking1.jpg 4 5
3 2-1 nosmoking.jpg 5 3-1 caution.jpg 4 2-2 nosmoking1.jpg 6 3-2 otherdangers.jpg 4 5 5.3 2 3 5 5 6 5 6, 5 caution.jpg 4 otherdangers.jpg 5 6 5.4 2 2 5
7 8 7 8 7 donotenter1.jpg 8 carsdonotenter.jpg 7 8 8 5.5 2 5 5 9 10 9 5-1 police.jpg 7 4-1 donotenter1.jpg 10 5-2 crewroom.jpg 8 4-2 carsdonotenter.jpg 9 10 police.jpg, crewroom.jpg
6. 1 CBIR 2 5 2 3 4 5 [3][4] 5 [1], DEWS 2008. [2] I. Tanaka and J. Suzuki, Web and Database Technologies, Proc. of ACM SIGMOD, pp. 10-22, 2010. [3] ( ) 2002 [4] Rayan Abdullah and Roger Hubner, SIGN, ICON and PICTOGRAM R.I.C 2006 [5] Ali Ridho Barakbah, Yasushi Kiyoki: A New Approach for Image Segmentation using Pillar-Kmeans Algorithm, International Journal of Signal Processing, Vol. 6, No. 2, pp. 82-87, WASET, 2010. [6] Ali Ridho Barakbah, Yasushi Kiyoki: An Image Search System with Analytical Functions for 3D Color Vector Quantization and Cluster-based Shape and Structure Features, Information Modelling and Knowledge Bases, Vol. XXI, pp. 169-187, IOS PRESS, March 2010.