Frequency analysis to facilitate the recognition of images

Content based image retrieval is an approach for facilitating the searching and browsing of large image collections. In this approach, image analysis is conducted on low level visual properties extracted from the image. We believed that in order to create an effective image retrieval system, visual perception must be taken into account. We conjectured that a technique which employs both M and P pathways would be more effective in the discrimination and search tasks of thumbnails. While the P pathway representations are easy to judge by observing color and shape, assessment of M pathway is possible through spatial frequency filtering. In order to validate this claim, a study was conducted to look at how manipulating spatial frequency composition of thumbnails affects visual search time. Thumbnails were sorted using frequency analysis and a user study measured recognition time, correctness of response, and number of repeated trials.