8.4. Representation and Indexing
We discussed the ultimate form of spatial data by grouping the data into object silhouettes, clusters of points, or point sets. In the next section, we leave the spatial domain to condense the pictorial information into feature values.
8.4.1. Grouping data
In content-based image retrieval, the image is often divided in parts before features are computed from each part. Partitionings of the image aim at obtaining more selective features by selecting pixels in a tradeoff against having more information in features when no subdivision of the image is used at all. We distinguish the following partitionings:
When searching for an object, it would be most advantageous to do a complete object segmentation first:
Strong ...
Get Emerging Topics in Computer Vision now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
