Abstract: Today humans live in the digital video age where everything we need, is available in terms of video information in the vast repositories online. There is a need to enable user devices and access existing vast store of video data in an easy manner. Although videos made day-to-day life easier and much more enjoyable due to the simplicity and flexibility that we get with the internet. In terms of video content retrieval all user must do, is to type in a search term and get back a relevant result. This process has its limitations in terms of time and speed of search tiny video is a search program that makes video retrieval more relevant in terms of the content within the video not regarding for the actual data tagged with it. It therefore promises greater accuracy in terms of relevant search information and is capable of cross-referencing an image or another sample video and not just text, to give a valid result to the user. This search algorithm is therefore an improvement on the current system and shows promise in terms of its accuracy. This paper examines the use of a unique search algorithm to improve video tagging and referencing given a large database of submitted content such as YouTube. We present our algorithm with 99% accuracy with database videos and speed of search increased 4 times than the existing search techniques.
Keywords: Tiny videos, Kernel, Video retrieval.