An efficient stochastic approach to groupwise non-rigid image registration

Abstract

The groupwise approach to non-rigid image registration, solving the dense correspondence problem, has recently been shown to be a useful tool in many applications, including medical imaging, automatic construction of statistical models of appearance and analysis of facial dynamics. Such an approach overcomes limitations of traditional pairwise methods but at a cost of having to search for the solution (optimal registration) in a space of much higher dimensionality which grows rapidly with the number of examples (images) being registered. Techniques to overcome this dimensionality problem have not been addressed sufficiently in the groupwise registration literature.In this paper, we propose a novel, fast and reliable, fully unsupervised stochastic algorithm to search for optimal groupwise dense correspondence in large sets of unmarked images. The efficiency of our approach stems from novel dimensionality reduction techniques specific to the problem of groupwise image registration and from comparative insensitivity of the adopted optimisation scheme (Simultaneous Perturbation Stochastic Approximation (SPSA)) to the high dimensionality of the search space. Additionally, our algorithm is formulated in way readily suited to implementation on graphics processing units (GPU).In evaluation of our method we show a high robustness and success rate, fast convergence on various types of test data, including facial images featuring large degrees of both inter-and intra-person variation, and show considerable improvement in terms of accuracy of solution and speed compared to traditional methods.

Learnt Real‐time Meshless Simulation

“…Registration involves the analysis of deformable structures in groups of images or surfaces and the construction of some statistical model of appearance or shape [DTT08]. A recent innovation in this field of non‐rigid image registration [SRM09] offers an efficient fully automatic approach that uses novel ideas in dimensionality reduction and high‐dimensional global optimisation to achieve high robustness and fast convergence on various types of imagery. Extending the ideas of [SRM09], we apply their method to align and register surfaces.…”

“…A recent innovation in this field of non‐rigid image registration [SRM09] offers an efficient fully automatic approach that uses novel ideas in dimensionality reduction and high‐dimensional global optimisation to achieve high robustness and fast convergence on various types of imagery. Extending the ideas of [SRM09], we apply their method to align and register surfaces. To achieve the highest possible accuracy of registration, we utilise the information contained in the textures (if available) rather than relying solely on the shape of the samples, see Section 3.1.…”

“…The textures are then registered using a groupwise image registration algorithm described in [SRM09] (but see also [TM04]). Thus a deformation map is computed for each texture which brings it into correspondence (densely) with a common reference frame.…”

“…Non‐rigid image and surface registration techniques, solving the dense correspondence problem, have emerged with applications in medical imaging, automatic construction of statistical models of appearance and analysis of facial dynamics. One family of approaches, groupwise registration [DTT08, SRM09], overcomes limitations of traditional pairwise methods and, therefore, allows for a very robust fully automated registration.…”

3D Stent Recovery from One X-Ray Projection

“…In order to reduce the complexity in our registration procedure, we employ a semi-simultaneous optimization framework based on the idea of Sidorov et al [9]. Instead of optimizing all parameters for all segments at once, we optimize the parameters of one segment for a certain number of iterations and then move to the next randomly chosen segment.…”