Image matching is one of the most important issues in object localization algorithms, while stable feature detection and
representation is a fundamental component of many image matching algorithms. SIFT algorithm has been identified as
the most resistant feature extraction method to common image deformations. In this paper, we use SSIFT (Simplified
Scale Invariant Feature Transform) to solve the problem of image matching in non-structured underwater environments.
Like SIFT, we construct a Gaussian pyramid and search for local peaks in a series of difference-of-Gaussian (DOG)
images; however, instead of using local square image patch to assign orientation and build 128-element vector, we apply
local circle image region and build only 12-element vector for each keypoint. The experiments have shown that SSIFT
are more robust to image rotation, and more compact than the standard SIFT representation. We also present fast
matching results using such descriptors for non-structured underwater objects.
|