Other Free Encyclopedias » Online Encyclopedia » Encyclopedia - Featured Articles » Contributed Topics from U-Z

Video Inpainting

motion visual processing restoration

Definition: Video inpainting refers to digital video restoration and video inpainting techniques should perform spatiotemporal restoration and adapt itself to the varying structural and motion characteristics of the visual data.

Digital inpainting plays a crucial role in digital video restoration. Digital processing of archived video data, transmission over best effort networks or wireless communication channels, and aggressive coding introduce visual impairments in video sequences. For example, channel fading during the wireless transmission of MPEG-coded videos can lead to packet loss which causes missed blocks in a received video. Furthermore, archived films and videos are exposed to chemical and physical elements as well as environmental conditions, which cause visual information loss and artifacts in the corresponding digital representation.

Motion video can be viewed as a three-dimensional (3-D) image signal or a time sequence of two-dimensional (2-D) images (frames). Such a visual input exhibits significant spatial and temporal correlation. Temporal restoration of motion video without spatial processing results in blurring of the structural information in the reconstructed video regardless of the motion or non-motion compensated nature of the temporal processing. On the other hand, processing each one of the video frames as still images and ignoring their temporal correlation produces strong motion artifacts. Therefore, a well designed video inpainting solution should perform spatiotemporal restoration and adapt itself to the varying structural and motion characteristics of the visual data.

In restoring archived films and videos, the presence of dirt, dust and scratches in the original medium results in speckle noise, random patches and sparkles of varying sizes and intensity. In the case of transmission over best effort type of networks, dropped macroblocks create missing data patches while transmission over erasure channels, such as the Gilbert-Elliot channel may introduce random errors at the pixel level. The spatial position of such visual impairments varies significantly as they are application dependent. It is therefore reasonable to assume that such impairments should be modeled and localized as temporal discontinuity. Through the employed motion compensation algorithms, this discontinuity is viewed as a spatial area in the actual frame which cannot be matched to a similar area in reference frames. Thus, an artifact detection procedure should precede the inpainting technique. The detector should be able to differentiate between structural content and visual artifacts. Such design characteristics can be obtained either through the heuristic threshold(s)-based detection methods as well as model-based approaches. After localizing the artifacts at the target frame, either a conventional image inpainting technique or a spatiotemporal inpainting solution can be used to fill-in the missing color, structural or textural information.

It should be noted that the utilization of an automated or semi-automated video inpainting techniques is essential for a cost-effective processing of the input. In most cases processing is performed in batch mode. Therefore, the development of high-performance parallel computational platform based video restoration techniques is of paramount importance.

Video Over Ip - Introduction, Video Requirements, Video Compression Standards, Session Setup, Video Distribution, Quality of Service (QoS) [next] [back] Video Delivery Over Wireless Multi-Hop Networks

User Comments

Your email address will be altered so spam harvesting bots can't read it easily.
Hide my email completely instead?

Cancel or