| Summary: | 碩士 === 國立臺灣大學 === 資訊工程學研究所 === 97 === Public and home safeties are important in human life. For the sake of security, it is necessary to have as many monitors of the surveillance system as possible to monitor specific regions. However, there is usually not sufficient bandwidth of networks to return all video streams, especially in narrowband networks.
To solve this issue, we propose a modified H.264 encoder, which uses little bandwidth to transfer more video streams as possible. The key concept is to compress all input video streams into one. Because we just know what happens in specific regions, it is not imperative to store high-quality video. The video should be small size instead of precise. We shrink four observed videos to one-quarter and combine them into one as long as the quality is distinct and smooth enough. This way, it is able to reduce the needed bandwidth and have acceptable quality. Still, each video has different frame rate, it is impossible to get samples in all regions when sampling. We modify the original encoder to improve encoding time when some regions get no sample. Finally, we encode the combined image by using modified H.264 encoder.
For evaluations, we compare the needed bandwidth of output generated by our system with that of traditional IP camera which is no scaling and combining on the input files, and the encoding time of modified encoder with that of original encoder. As confirmed by performance evaluations, the proposed modified H.264 encoder with limited hardware cost can achieve excellent performance in term of the bandwidth of transmissions and encoding time.
|
|---|
