Adaptive Frame Features Based Video Footage Retrieval System

Due to high dimensionality of video data, video mining becomes tedious task for any search engine. It requires lot of computation on account of multiple and nested loops and that too at high speed. This makes the computational task very expensive in terms of programming performance like speed of operation and code size etc. However, if the video sequences are stored based on contents like color, texture, or events, then the video mining may be speed up to a great extent. In the presented work, the color based and identical things (or repetitive things) in continuous frames based, video mining is proposed for speedy search at fair accuracy. In the proposed work, recorded video is scanned based on query contents like color (fire) by decomposing the video into frames. A video sequence is a collection of no. of 2-d images played over a time dimension. The 2-d frame are analysed for color based features contents and integrated with period of time so that complete video is scanned for the presence of color in the video sequence. The same concept may be applied for other features like shape, motion artifact. The video frames are decomposed into frames using time based images frames extractions. The time interval may be decided based on gravity of the feature like color in case of fire and motion.