Content-Based Video Retrieval Based on Multifaceted Feature Extraction Using Statistical Methods

Abstract

For content-dependent video recovery, this paper proposes a multifaceted feature extraction approach based on color string extraction, local texture characteristics, and amount of Absolute Difference, which extracts the features such as color components , texture components, and motion behavior. The color string function is derived using the string duration count, and the color histogram approach is used for detecting scene shifts. Based on scene shift recognition, the key-frames are retrieved. Based on the color string and local texture functions, the color and texture characteristics are derived from the key-frame. The Amount of Absolute Difference is measured, and the motion function of a video is derived using it. To compare the function vectors of the question and main frames of the video to be recovered, the Jeffrey's Divergence measure is implemented. The suggested approach has the greatest capture of both spatial and temporal aspects, and the recovery efficiency is increased by utilizing multifaceted features. The system suggested outperforms the current approaches.