Abstract
Similarity measures are very crucial especially in the field of information retrieval. Thus, various distance/similarity measures were proposed throughout the literature. In the video retrieval field, videos are represented as multi-dimensional features vector. Once this features vector is extracted from video shots; the retrieval task is primarily performed based on the measurement of similarity between respective videos’ feature vectors. Moreover, the retrieval quality could be greatly improved with careful distance measure selection. This paper presents an extensive analysis regarding the most commonly used video retrieval similarity measures. The results are consolidated with a multifaceted analysis, i.e. multiple challenging video datasets, retrieval curves and confusion matrices. The major contribution of this paper is investigating the effectiveness of the common similarity measures from a video retrieval perspective. This would give the field researchers the required knowledge to select the most suitable distance measure for their video retrieval research work.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call