This thesis presents several approaches for preserving the most important image features when compressing images or videos at relatively low bitrates in a standard-compatible manner. The proposed schemes exploit the characteristics of both feature detection and description. Additionally, a new rate-distortion model and a novel rate control framework are proposed for video encoding. Finally, the schemes are further improved by explicitly encoding the original keypoints. The experiments are performed based on the JPEG, H.264/AVC, and H.265/HEVC standards, and the improved feature-matching performance is demonstrated.
«
This thesis presents several approaches for preserving the most important image features when compressing images or videos at relatively low bitrates in a standard-compatible manner. The proposed schemes exploit the characteristics of both feature detection and description. Additionally, a new rate-distortion model and a novel rate control framework are proposed for video encoding. Finally, the schemes are further improved by explicitly encoding the original keypoints. The experiments are perfor...
»