SURF (Speeded Up Robust Features) is a robust image detector & descriptor, first presented by Herbert Bay et al. in 2006, that can be used in computer vision tasks like object recognition or 3D reconstruction. It is partly inspired by the SIFT descriptor. The standard version of SURF is several times faster than SIFT and claimed by its authors to be more robust against different image transformations than SIFT. SURF is based on sums of approximated 2D Haar wavelet responses and makes an efficient use of integral images. As basic image features it uses a Haar wavelet approximation of the determinant of Hessian blob detector (WIKIPEDIA).
The performance gain in the SURF Implementation is due to use of integral images (See my post on Integral Images)
In constructing the scale space, Lowe proposed a difference of Gaussian to approximate the Laplacian of Gaussian.The images were blurred and subsampled to produce a scale space. After which the difference of gaussian was used to estimate the laplacian gaussian.
in SURF this was done using a box filter representation of a kernel. This kernel was used to produce the pyramid and kept the original image unchanged while changing the filter size.
TO DO: Add more of the details
No comments:
Post a Comment