34240 - Data Compression
and Image Communication
3D-DCT Video Coding for Static
Camera Streaming Applications
Brajato Giovanni, DTU Department of Photonics
s160975@student.dtu.dk
Abstract
In this report a video compression technique based on a 3-
dimensional discrete cosine transform is proposed,
implemented and analysed. The whole system
encoder/decoder is extensively described, from the theoretical
aspects of the main coding blocks, to a functional
implementation and testing phase using MATLAB®. The
performances of the system are evaluated via distortion/rate
measures on different test videos and by comparing it with
others existing compression schemes. Additional results include
quality assessment and entropy estimation.
INTRODUCTION
Nowadays, video streaming and broadcast over internet is becoming very popular, both in social networks
and streaming platforms [1] [2]. This feature uses a lot of the bandwidth available [3] [4] therefore
efficient coding algorithms are desirable. Not only the encoder has to be fast and reliable, but it must
compress the original data as much as possible by preserving an acceptable quality, to not use the whole
bandwidth available.
This is true especially when the amount of data accumulates continuously over time, as it happens when
surveillance static cameras stream over the network 24h. This kind of video are characterized by a lot of
temporal redundancy which may be exploited using a transformation in the temporal domain (inter frame
correlation). One way to achieve it is through predictive coding [5]. However, to reduce the complexity of
those algorithms [6], one may use a transform coding [5] to compress as eliminate all frame redundancies
