CODEWITHSUNDEEP
matlabopencvimage-processingimage-segmentationobject-detection
Ashutosh Mishra
Ashutosh Mishra

Reputation: 107

Why superpixel should adhere well to image boundaries?

I am reading a paper from EPFL SLIC-Superpixel comparision to state-of-the-art which says that superpixel should adhere to image boundaries. What is the need of adhering to image boundary of the image ?

Upvotes: 3

Views: 285

Answers (1)

Jonas
Jonas

Reputation: 74940

A superpixel is an image region containing similar pixels. An image boundary, or edge, is a boundary between different parts of the image (not to be confused with the outside border of the image). As a consequence, a good superpixel algorithm will have all edges in the image covered by superpixel boundaries, because otherwise, this would mean that there is an edge going through a superpixel, i.e. the superpixel would span two different image regions, and that is not compatible with the superpixel definition.

Of course, as superpixels become larger and larger, and allow including pixels that are more and more different from one another, superpixels will start extending past weak edges in the image. Strong edges will, however, still be covered by superpixel boundaries.

Upvotes: 4

Related Questions