Semantic retrieval from video databases is becoming a very important resear
ch topic in the area of multimedia. This kind of tasks require the developm
ent of video data representation models which include the relationships bet
ween low-level visual cues and the semantic concepts inferred from them. Th
is paper presents a work based on semiotic studies that includes the extrac
tion of simple visual features from commercials and a statistical analysis
of them and their relationships with high-level semantic terms. Well-known
algorithms have been implemented and enhanced for feature extraction, as we
ll as a novel probabilistic approach to color naming. The statistical analy
sis consists of finding correlations between variables, as well as the dime
nsions in feature space that best explain the variance of the data set. Som
e interesting conclusions are reached at the end of the work about how comm
ercials are grouped in feature space with respect to different levels of se
mantics. (C) 2000 Academic Press.