In-image similarity-based image structure descriptor and its applications

Research Project

Project/Area Number	15K00156
Research Category	Grant-in-Aid for Scientific Research (C)
Allocation Type	Multi-year Fund
Section	一般
Research Field	Multimedia database
Research Institution	Tokyo Metropolitan University
Principal Investigator	Fujiyoshi Masaaki 首都大学東京, 学術情報基盤センター, 准教授 (20336522)
Project Period (FY)	2015-04-01 – 2019-03-31
Project Status	Completed (Fiscal Year 2018)
Budget Amount *help	¥4,680,000 (Direct Cost: ¥3,600,000、Indirect Cost: ¥1,080,000) Fiscal Year 2017: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000) Fiscal Year 2016: ¥1,170,000 (Direct Cost: ¥900,000、Indirect Cost: ¥270,000) Fiscal Year 2015: ¥2,080,000 (Direct Cost: ¥1,600,000、Indirect Cost: ¥480,000)
Keywords	類似性 / 画像内類似性 / 情報通信システム / 画像処理
Outline of Final Research Achievements	This study aimed at developing the image structure descriptor based on relations among similar areas in single image. First, several techniques for computing the similarity between two image pieces were investigated by applying techniques to various images. Unfortunately, the best technique for this study has not been decided because the techniques varied in the performance due to the size and content of images. Computing the similarities between any areas requires a huge computational cost, even if areas are in single image. Then, the similarities were calculated between feature points taken out by several techniques. Again, the best technique combination could not be figured out because the same reason. Besides these results, a method for automatic classification of photography composition has developed based on the results, and the method has been published.
Academic Significance and Societal Importance of the Research Achievements	目標とした構造記述子は，画像の不正編集検出などに有用であると考えられる．一方，研究遂行中に着想を得，関連研究として成果を公表した，風景写真構図の自動推定は，大量のディジタル写真を日々撮影し，データを蓄積する現在，保存あるいは共有などするに値する好ましい写真を自動的に選別する一助となり，利用者の負担を軽減する．従来，顔写真を対象とした技術しかなかったが，その範囲を風景にも広げている．