Data embedding using predictive syntax elements in scalable coded video / Pang Lie Lin

Pang , Lie Lin (2021) Data embedding using predictive syntax elements in scalable coded video / Pang Lie Lin. PhD thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (185Kb)
	PDF (Thesis PhD) Download (2051Kb)

Abstract

With the rapid advancement in digital technologies, video has emerged as one of the most effective and popular communication media. While videos are increasingly adopted for various purposes including entertainment, dashboard camera recording, and recently education, many issues related to video arise, including unauthorized or illegal use of video content, tampered video content, and packet loss due to network congestion or bandwidth fluctuation. As a result, various proposals are put forward to manage videos, and one of them is data embedding. Essentially, data embedding inserts data into a video to serve a specific purpose, including proof of ownership via watermark, covert communication in steganography, as well as error concealment and authentication via fragile watermark. This work focuses on embedding data into video coded in the format of Scalable High Efficiency Video Coding (SHVC), which plays an important role in adaptive video streaming applications. The scalability feature is essential, notably when considering the heterogeneity in transmission and decoding devices. However, the underlying architecture of this scalable video coding standard differs from the previous scalable video coding standards. A study is therefore required to explore the possibility to embed data into SHVC encoded video. Specifically, the predictive syntax elements in the SHVC compressed domain are analyzed. To increase embedding capacity, payload bits are embedded in all scalable coded layers. Conceptually, coding a video frame using more (smaller) prediction blocks leads to more embedding opportunities due to the availability of the syntax elements. Therefore, an adjustable control parameter is introduced to guide the quad-tree partitioning process to permit more split blocks. The proposed technique achieves encouraging payload with slight bit rate overhead. When a predictive syntax element is utilized for data embedding, the encoder calculates and encodes the prediction error caused by data embedding as a portion of the prediction residual and reconstruct the video as similar as possible to the original video. Hence, the manipulation of predictive syntax elements will not cause drift-error. In view of this, both intra- and inter-coded blocks are jointly utilized for data embedding without compromising the perceptual quality. During encoding, the prediction modes and partitioning depth are assessed by rate distortion optimizer, and only the one with the best rate distortion cost is adopted for coding. A data embedding framework named CUPSEED, which automatically selects different prediction elements for data embedding, is then proposed to manage multiple data embedding venues and coding layers. The proposed framework outperforms existing data embedding techniques in the SHVC compressed domain in term of embedding capacity. It manages to achieve higher payload while preserving the perceptual quality with minimal bit rate variation.

Actions (For repository staff only : Login required)