Rotation-invariant wavelet-based matching of local image features, with enhanced tolerance to shifts in location and scale.
Nick Kingsbury (University of Cambridge)
COMPUTER VISION AND ROBOTICS SERIESDATE: 2011-04-07
TIME: 16:00:00 - 17:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
I shall describe a technique for using dual-tree complex wavelets to obtain rich feature descriptors of keypoints in images. The main aim has been to develop a method for retaining the full phase and amplitude information from the complex wavelet coefficients at each scale, while presenting the feature descriptors in a Fourier-domain form that allows for efficient correlation at arbitrary rotations between the candidate and reference image patches. The feature descriptors are known as Polar-Matching matrices. Recently, we have modified our previouly proposed approach so that it can be more resilient to errors in keypoint location and scale. These multi-scale feature descriptors are potentially useful for object detection, recognition, classification and tracking in images and video. This work was done in collaboration with James Nelson, now at University College London. I will also briefly discuss the sort of areas that I shall be interested in pursuing while I am in Canberra for the next 6 weeks, which include optimisation of feature descriptors to work with classifiers and machine learning systems to produce improved vision systems.
BIO:
Nick Kingsbury is Professor of Signal Processing at the University of Cambridge, Department of Engineering. He has worked in the areas of digital communications, audio analysis and coding, and image processing. He has developed the dual-tree complex wavelet transform and is especially interested in the application of complex wavelets and related multiscale and multiresolution methods to the analysis of images and 3-D datasets.
