Abstract
The main difficulty in implementing security to real-time videos is the processing time. Processing time is considerably high while ensuring security using single-core processors. This paper is analyzing the processing capability of Multicore CPUs and GPUs to facilitate a secure video transmission. Six different techniques using spatial and frequency domains are implemented using these systems. Adjacent frames are secured using different techniques to improve security. The processed video is transmitted over LAN to a neighboring system to see whether a real-time reproduction is possible or not. The paper also compares the processing time for real-time as well as stored videos, with varying resolutions. In the case of real-time video security, the GPU system was able to transmit 23 frames per second while the single-core CPU system was able to transmit only 2 frames per second. Multicore CPU system with 8 cores was able to transmit 8 frames per second. The resolution of the video transmitted was 320x240. When just security techniques were applied (video not transmitted) on a stored video of resolution 640x480, the performance of the GPU system was 38.3 times better than single-core CPU system and 7.7 times better than multicore CPU system.
Jerin Geogy George
Research Scholar
My research interests include meta-optics, computational electromagnetics, dielectric metasurfaces, deep learning