Yutaka Hatakeyama

Image retrieval based on similarity score fusion from feature similarity ranking lists

An image similarity method based on the fusion of similarity scores of feature similarity ranking lists is proposed. It takes an advantage of combining the similarity value scores of all feature types representing the image content by means of different integration algorithms when computing the image similarity. Three fusion algorithms for the purpose of fusing image feature similarity scores from the feature similarity ranking lists are proposed. Image retrieval experimental results of the evaluation on four general purpose image databases with 4,444 images classified into 150 semantic categories reveal that a proposed method results in the best overall retrieval performance in comparison to the methods employing single feature similarity lists when determining image similarity with an average retrieval precision higher about 15%. Compared to two well-known image retrieval system, SIMPLicity and WBIIS, the proposed method brings an increase of 4% and 27% respectively in average retrieval precision. The proposed method based on multiple criteria thus provides better approximation of the users similarity criteria when modeling image similarity.

Color restoration algorithm for dynamic images under multiple luminance conditions using correction vectors

An algorithm for color restoration under multiple luminance conditions is proposed. It automatically produces correction vectors to restore the color information in the L^*a^*b^* color metric space, using color values of a target object within the well-illuminated region in a given dynamic image. The use of the correction vectors provides better image quality than that obtained by the restoration algorithm using color change vectors. An experiment is done with two real dynamic images, where a walking person in a building is observed, to evaluate the performance of the proposed algorithm in terms of color-difference. The experimental results show that the restored image by the proposed algorithm decreases the color-difference by 30% compared to the restoration algorithm using color change vectors. The proposed algorithm presents the foundation to identify the person captured by a practical security system using a low cost CCD camera.

Detection Algorithm for Color Image by Multiple Surveillance Camera under Low Illumination Based-on Fuzzy Corresponding Map

An objects detection algorithm for color dynamic images from two cameras is proposed for a real surveillance system under low illumination. It provides automatic calculation of a Fuzzy Corresponding Map and color similarity for lower luminance conditions, which detects small chromatic regions in CCD camera images under lower illumination. Experimental detection results for two dynamic images from real surveillance cameras in a downtown area in Japan under low luminance conditions show that the proposed algorithm has 15% improved accuracy compared with the independent detection algorithm in the same false alarm rate, which implementability for severe surveillance situation is discussed. The proposed algorithm is being considered for use in a low cost surveillance system at a relatively poor security downtown (shopping mall) area in Japan.

Automatic Image Annotation based-on Rough Set Theory with Visual Keys

For automatic image annotation, a method based on rough sets with visual keys is proposed. Using rough set theory the method constructs decision rules about each visual key used for image indexing and about keywords from training set of already annotated images. Then target image is annotated according to constructed decision rules about visual keys which the target image is indexed by. The method is evaluated with training sets of 900 images and with test sets of 100 images on 1,000 manually annotated images in COREL database. Experiments show that recall rates tend to rise easily compared with precision rates on image retrieval with query-by-keywords

Cross-reference Detection Algorithm for the Real Surveillance Systems Based-on Fuzzy Corresponding Map

A detection algorithm with color information for two dynamic images in a real surveillance system is proposed. It considers input region of frame as a detected region or the region behind other objects based on fuzzy corresponding map which describes common regions between input two dynamic images. Detection experimental results for the real surveillance situation show that the proposed algorithm improves 30% of accuracy compared with the independent detection algorithm. The algorithm will be installed in a basis unit for the real surveillance systems.

Instance-based location estimation algorithm for a pedestrian in multiple color dynamic images

Location estimation algorithm for a pedestrian is proposed for the real surveillance system based on color instances with color dynamic images under low illumination, where the proposed color instances consist of color-difference, moving possibility region, and previous detection objects in edge region using time series data. It provides useful detection result for too low illuminated situation. Experimental results for dynamic image taken under low illumination in streets show that detected frames with the proposed algorithm increase by 20% compared with detection result without color instances. The proposed algorithm is under consideration for use in a relatively poor security downtown area in Japan.