TESTBED DEVELOPMENT OF INTEGRATED VIDEO CODEC BASED ON CNN AND CONVENTIONAL: DVC ENCODER, DPCM DECODER, AND H.264

• Main Author: Bimo Prayogo, Aloysius
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/64584
• Institution: Institut Teknologi Bandung

Video data nowadays are used a lot in every sector, not only for saving memories, but tend to be as important as sharing information, communication tools, and another kind of important activites. During pandemic era, people discussed and gathered using digital platform with facility such as realtime communicaiton using video and audio. As the increasing amount of transmitted video data, all the supporting systems and infrastructures also have to be improved, more reliable, and effective enough for transmitting data. Data transmission often use compression mechanism to decrease bandwidth needed in one transmission. One of compression types is lossy, the one that will reduce a little of video size and quality when it is arrived at the receiver due to quantization phase. This system is frequently called as video codecs that is generally divided into encoder for video compression and decoder for generating approximattely the same video for playback. Some of conventional compression frameworks today are H.264, VP8, and RV40. Video codecs framework improvement can be done by utilizing AI for representation of non-linear value ability based on motion-compensatation parts and residual transform. Deep Video Coding (DVC) framework utilize motion information from optical flow mechanism. DVC has adaptive quantization feature that can be different to each processed video information to reduce bit rate variable parameter. Development process does not consider time consumption variable, instead considering on ratio compression parameter optimal quality. System can be accessed with simple Graphical User Interface that takes input of video, then dividing it into couple of frames that will be inferred one by one and final result of compressed video with quality and ratio information.