## Image Registration with Automatic Computation of Gradients Release 1.0 (2008)

### BibTeX

@MISC{Kahn08imageregistration,

author = {E. G. Kahn and L. H. Staib},

title = {Image Registration with Automatic Computation of Gradients Release 1.0},

year = {2008}

}

### OpenURL

### Abstract

Many image registration algorithms are formulated as optimization problems with a gradient descent based solver, One difficulty with designing and implementing such methods is in the implementation of the gradient computation. This process can be time-consuming and error-prone. In addition some functions do not have gradients that can be expressed in symbolic form. Automatic differentiation is useful for computing gradients of complicated objective functions. It moves the burden of computing gradients from the programmer to the computer. So far, AD has not been exploited for use in image registration. This paper describes a software library the authors have developed to automate the process of computing gradients of registration objective functions. This can alleviate the job of registration designers somewhat and potentially make it easier to design better registration algorithms.

### Citations

492 |
der Vorst. Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods, 2nd Edition
- Barrett, Berry, et al.
- 1994
(Show Context)
Citation Context ... • itk::KernelTransform -> itkek::ADRevolveRBFTransform Our new version of itk::KernelTransform is for the special case of a thin plate spline radial basis function transform in 3D which uses a GMRES =-=[1]-=- linear system iteration for solving for the coefficients. We called this new class ADRevolveRBFTransform. Note that it not yet possible in ITK to compute gradients of metrics that use a KernelTransfo... |

300 |
Evaluating Derivatives: Principles and Techniques of Algorithmic Differentiation
- Griewank
- 2000
(Show Context)
Citation Context ... if we could avoid having to write these functions and yet still compute an exact gradient within a reasonable amount of time. Fortunately, there is a solution known as Automatic Differentiation (AD) =-=[3]-=-. Automatic differentiation is a method for overcoming such problems by having the computer automatically compute gradients in a fast and efficient way. In addition, because automatic differentiation ... |

114 | Achieving logarithmic growth of temporal and spatial complexity in reverse communication
- Griewank
- 1992
(Show Context)
Citation Context ...s not scale well to large problems due to large memory requirements, another scheme known as “checkpointing” can be combined with AD to allow the gradient evaluation of functions of virtually any size=-=[2, 5]-=-. To date, the use of AD with checkpointing does not appear to have become popular in image registration problems. In this paper, we describe the integration of automatic differentiation with ITK. Thi... |

23 |
The ITK software guide: The insight segmentation and registration toolkit (version 1.4
- Ibanez, Schroeder, et al.
- 2003
(Show Context)
Citation Context ...the checkpointing algorithm. Unfortunately, these two packages had not previously been merged together (to our knowledge). Therefore it was necessary to combine them together before using them in ITK =-=[6]-=-. The result is a library called adcheck located in the directory src/adcheck. This library is completely independent of ITK and can be used to compute gradients of any scalar objective function. In o... |

18 |
Algorithm 755: ADOL-C: a package for the automatic dierentiation of algorithms written in C/C
- Griewank, Juedes, et al.
- 1996
(Show Context)
Citation Context ...utomatic differentiation with ITK. This paper assumes the reader has a good understanding of AD as a full discussion of these issues is beyond the scope if this work. For more details, see References =-=[3, 4, 5, 7]-=-. This paper is organized as follows. In Section 2, we describe how we integrated AD with ITK’s registration framework. In Section 3 we describe some experiments conducted with our method and Section ... |

16 |
2000 Algorithm 799: revolve: an implementation of checkpointing for the reverse or adjoint mode of computational differentiation
- Griewank, Walther
(Show Context)
Citation Context ...s not scale well to large problems due to large memory requirements, another scheme known as “checkpointing” can be combined with AD to allow the gradient evaluation of functions of virtually any size=-=[2, 5]-=-. To date, the use of AD with checkpointing does not appear to have become popular in image registration problems. In this paper, we describe the integration of automatic differentiation with ITK. Thi... |

1 | Computational Strategies for Meshfree Nonrigid Registration
- Kahn
- 2006
(Show Context)
Citation Context ...utomatic differentiation with ITK. This paper assumes the reader has a good understanding of AD as a full discussion of these issues is beyond the scope if this work. For more details, see References =-=[3, 4, 5, 7]-=-. This paper is organized as follows. In Section 2, we describe how we integrated AD with ITK’s registration framework. In Section 3 we describe some experiments conducted with our method and Section ... |