Volume 19, No. 2, 2022
Visual Saliency Estimation Using Combined Feature Of PDDL And MRF
Vinay C Warad and Dr. Ruksar Fatima
Abstract
-Last few decades has seen the big growth in visual saliency due to its usability in the real time scenario. In past several models have been proposed, despite of research variety the existing model fails to match the human level accuracy. In this research work, we have proposed a technique named as SEIPM –video encoding, in here the video size is compressed without degrading any quality in the video. Moreover, SEIPM is based on the saliency prediction of spatial and temporal features; these features are combined by PDDL (Processing Division of Description Layer) and MRF (Markov Random Field). MRF is used for analyzing the saliency frame to recognize movement and motions in the frame and PDDL is used for the feature extraction which are very much eye attracting. Once the region of movement and the particular region, which attracts the eye, is recognized, then encoding is done from the previous frame only and hence we tend to achieve the quality video. Moreover, SEIPM-video encoding is evaluated by considering the standard dataset used in existing protocol by considering the various performance metric and compared with the various state-of-art technique and our models outperforms the other state-of-art technique.
Pages: 3359-3370
Keywords: The widespread utilization of inexpensive and portable video capture devices, such as camcorders and mobile phones, coupled with the surge in surveillance cameras, has led to a remarkable increase for data.