TY - GEN
T1 - Psychophysical evaluation for a qualitative semantic image categorisation and retrieval approach
AU - Ul-Qayyum, Zia
AU - Klippel, Alexander
AU - Cohn, A.G.
PY - 2010
Y1 - 2010
N2 - This paper details the behavioral evaluation of a qualitative image categorisation and retrieval approach using semantic features of images. Content based image retrieval and classification systems are highly active research areas and a cognitively plausible image description can improve effectiveness of such systems. While most approaches focus on low level image feature in order to classify images, humans, while certainly relying on some aspects of low level features, also apply high-level classifications. These high-level classification are often qualitative in nature and we have implemented a qualitative image categorisation and retrieval framework to account for human cognitive principles. While the dataset, i.e. the image database that was used for classification and retrieval purposes contained images that where annotated and therefore provided some ground truth for assessing the validity of the algorithm, we decided to add an additional behavioral evaluation step: Participants performed similarity ratings on a carefully chosen subset of picture implemented as a grouping task. Instead of using a predefined number of categories, participants could make their own choice on a) how many groups they thought were appropriate and b) which icons/images belong into these groups. The results show that the overall underlying conceptual structure created by the participants corresponds well to the classification provided through the algorithm.
AB - This paper details the behavioral evaluation of a qualitative image categorisation and retrieval approach using semantic features of images. Content based image retrieval and classification systems are highly active research areas and a cognitively plausible image description can improve effectiveness of such systems. While most approaches focus on low level image feature in order to classify images, humans, while certainly relying on some aspects of low level features, also apply high-level classifications. These high-level classification are often qualitative in nature and we have implemented a qualitative image categorisation and retrieval framework to account for human cognitive principles. While the dataset, i.e. the image database that was used for classification and retrieval purposes contained images that where annotated and therefore provided some ground truth for assessing the validity of the algorithm, we decided to add an additional behavioral evaluation step: Participants performed similarity ratings on a carefully chosen subset of picture implemented as a grouping task. Instead of using a predefined number of categories, participants could make their own choice on a) how many groups they thought were appropriate and b) which icons/images belong into these groups. The results show that the overall underlying conceptual structure created by the participants corresponds well to the classification provided through the algorithm.
KW - image categorisation and retrieval
KW - psychophysical evaluation
KW - qualitative similarity
KW - Qualitative spatial representation
U2 - 10.1007/978-3-642-13033-5_33
DO - 10.1007/978-3-642-13033-5_33
M3 - Conference paper
AN - SCOPUS:79551507990
SN - 9783642130328
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 321
EP - 330
BT - Trends in Applied Intelligent Systems - 23rd International Conference on Industrial Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2010, Proceedings
T2 - 23rd International Conference on Industrial Engineering and Other Applications of Applied Intelligence Systems, IEA/AIE 2010
Y2 - 1 June 2010 through 4 June 2010
ER -