Technique | Segment DSs (VideoSegment DS, MovingRegion DS, and StillRegion DS) |
---|---|
Document | MPEG-7 Committee Draft MDS, see Segment DSs (VideoSegment DS, MovingRegion DS, and StillRegion DS) |
Name | Ana Belen Benitez, Columbia University |
Contact | ana@ee.columbia.edu |
Type | Application |
External Libraries | None |
Related Ds/DSs | Visual Descriptors and Description Schemes, SegmentRelation DSs, Segment attributes |
Used Ds/DSs | None |
Input | The input file contains a list of images with temporal and spatial information for a set of segments. Segments can be video segments (VideoSegment), still regions (StillRegion), or moving regions (MovingRegion) that can have associated any number of frames, frames + masks, and frame + masks, respectively. Each image contains that information as follows. The x-coordinate of the image represents the frame number; the y-coordinate represents the segment id; and the gray scale information provides the mask id for the segment in that frame. The location of the original frames and masks are obtai ned as follows: "[input image].dir/f_[frame number].ras" "[input image].dir/f_[mask id][frame number].ras" respectively. Together with this image, each input has a text file describing the hierarchical relationship between these segments, whose location is "[input image].dir/text.txt" and whose format is the following: [Segment id] [Segment type] [Number of children] [children Id's] [Decomposition type] [gaps?] [overlaps?] [Number of annotations] [Annotation's] where the decomposition and annotation information are optional. An example of a well formed input file with three still regions follows: 1 StillRegion 2 2 3 spatial no yes 1 elephant 2 StillRegion 0 3 StillRegion 0 |
Extraction | Yes |
Client Appl | Search & RetrievalSummary This code generates descriptions of segment trees. This code also allows to retrieve the segment descriptions based on the the minimum distance between pairs of segments in the segment trees. |
Strong Points | - |
Limitations | None |
Parameters | None |