The object-based MPEG4 standard is being increasingly adopted for its applications in the internet, wireless communication and video editing space. It reduces the bit-rate during transmission and simplifies the video-editing process in multimedia applications by allowing for scaling, rotation, and translation to meet editing requirements. In addition, it can also transmit only objects (without the background) to meet dynamic network bandwidth requirements.
However, these characteristics also make the object-based MPEG4 standard hard to secure from malicious attackers, as they can easily access the video object (VO) and modify or copy it from one video to another. While DRM protected content ensures safe transmission of videos, forensic watermarking schemes are needed to provide security once the video reaches the user device. This has made content owners and distributors migrate from the traditionally used frame-based watermarking approach to an object-based approach. In such a situation, an object is the unit for acquisition, editing, indexing, retrieval, coding, and distribution. While MPEG4 helps in better management and protection for media, the object functionalities are very simple, such as cutting and pasting. Hence, there is a need for object-based watermarking schemes in the MPEG4 environment.
The requirements for an object-based video watermarking scheme are different from those for a frame-based video authentication system, and they are usually application specific. In an object-based watermarking system, every object in the video should be processed, signed, and authenticated separately. Usually the video file is first segmented into foreground (objects) and background, which are then compressed individually for transmission. The watermark is then securely generated by combining the features extracted from both the segments. A secure link between the objects and background can then be established by embedding the watermark into the foreground objects. The authenticity of the video can be established by correlating the extracted watermark from the object and the extracted feature from both the received object and background.
The VO and background information can be easily modified or replaced by another object during malicious attacks (intentional distortions). At the user end the decompressed objects may be scaled, rotated, or translated to interact with end users (incidental distortions). An efficient object-based watermarking solution should, hence, be robust against incidental distortions while being able to detect intentional distortions for specific video requirements.
