A mean field annealing approach to accurate free form shape matching

Liu, Yonghuai

doi:10.1016/j.patcog.2006.11.025

Cited by 25 publications

(17 citation statements)

References 39 publications

(113 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[65] approximates the QAP by linearising the cost function via Taylor series and using two-way constraints to convert the QAP into a continuous search. Inspired by this, graduated assignment (e.g., [35], [59], [60]) and mean field annealing [13] techniques are further developments with additional constraints. These techniques use a rigid transformation instead of the quadratic term f ijdφ(i)φ(j) to model the pairwise relationship of correspondences (facilities & locations in QAP).…”

Section: Observations and Discussionmentioning

confidence: 99%

“…Our survey tries NonRigid Focus [10] surface registration for medical imaging [11] systematic breakdown of ICP and its variants [12] registration and fusion of range images [13] comparison of several Improved ICPs [14] comparison of quadratic approximants and ICP [15] coarse vs fine, pairwise vs multi-view alignments [16] techniques for shape correspondence to connect rigid and non-rigid registration, compare and contrast the novel techniques used, and discuss the underlying background to help readers appreciate the developments in the field. [10], [16] cover both rigid and non-rigid registration.…”

Section: Comparsion To Other Surveysmentioning

confidence: 99%

“…From an optimization point of view, this helps to convexify the error landscape of the objective function. [13] uses a similar argument involving equalization of confidence weightings based on mean field variables. Maximizing Correspondences: [65] uses the term − i j M ij to maximize correspondences and bias the objective function towards more matches (i.e.…”

Section: Regularization Constraintsmentioning

confidence: 99%

See 2 more Smart Citations

Registration of 3D Point Clouds and Meshes: A Survey from Rigid to Nonrigid

Tam

Cheng

Lai

et al. 2013

IEEE Trans. Visual. Comput. Graphics

575

268

View full text Add to dashboard Cite

Abstract-3D surface registration transforms multiple 3D datasets into the same coordinate system so as to align overlapping components of these sets. Recent surveys have covered different aspects of either rigid or non-rigid registration, but seldom discuss them as a whole. Our study serves two purposes: (i) to give a comprehensive survey of both types of registration, focusing on 3D point clouds and meshes, and (ii) to provide a better understanding of registration from the perspective of data fitting. Registration is closely related to data fitting in that it comprises three core interwoven components: model selection, correspondences & constraints and optimization. Study of these components (i) provides a basis for comparison of the novelties of different techniques, (ii) reveals the similarity of rigid and non-rigid registration in terms of problem representations, and (iii) shows how over-fitting arises in non-rigid registration and the reasons for increasing interest in intrinsic techniques. We further summarise some practical issues of registration which include initializations and evaluations, and discuss some of our own observations, insights and foreseeable research trends.

show abstract

Section: Observations and Discussionmentioning

confidence: 99%

Section: Comparsion To Other Surveysmentioning

confidence: 99%

Section: Regularization Constraintsmentioning

confidence: 99%

See 1 more Smart Citation

Registration of 3D Point Clouds and Meshes: A Survey from Rigid to Nonrigid

Tam

Cheng

Lai

et al. 2013

IEEE Trans. Visual. Comput. Graphics

575

268

View full text Add to dashboard Cite

show abstract

“…Model morphing and deformable shape registration [7]- [9] are commonplace in medical computational modelling [10]- [11]- [13]- [14]- [15] and computer vision [12]- [24]- [25]- [26]- [27].…”

Section: A Introduction To the Methodologymentioning

confidence: 99%

A Centerline-Based Model Morphing Algorithm for Patient-Specific Finite Element Modeling of the Left Ventricle

Behdadfar

Navarro

Sundnes

et al. 2018

IEEE Trans. Biomed. Eng.

View full text Add to dashboard Cite

Abstract-Goal: Hexahedral automatic model generation is a recurrent problem in computer vision and computational biomechanics. It may even become a challenging problem when one wants to develop a patient-specific finite-element (FE) model of the left ventricle (LV), particularly when only low resolution images are available. In the present study, a fast and efficient algorithm is presented and tested to address such a situation. Methods: A template FE hexahedral model was created for a LV geometry using a General Electric (GE) ultrasound (US) system. A system of centerline was considered for this LV mesh. Then, the nodes located over the endocardial and epicardial surfaces are respectively projected from this centerline onto the actual endocardial and epicardial surfaces reconstructed from a patient's US data. Finally, the position of the internal nodes is derived by finding the deformations with minimal elastic energy. This approach was applied to eight patients suffering from congestive heart disease. A FE analysis was performed to derive the stress induced in the LV tissue by diastolic blood pressure on each of them. Results: Our model morphing algorithm was applied successfully and the obtained meshes showed only marginal mismatches when compared to the corresponding US geometries. The diastolic FE analyses were successfully performed in seven patients to derive the distribution of principal stresses. Conclusion: The original model morphing algorithm is fast and robust with low computational cost. Significance: This low cost model morphing algorithm may be highly beneficial for future patient-specific reduced-order modelling of the LV with potential application to other crucial organs.

show abstract

“…While a complete combination of points in the overlapping free form shapes are all considered in [9,6,10] as tentative correspondences, the closest points are considered instead in [17,16] for computational efficiency. To estimate the probabilities that these tentative correspondences represent real ones, the entropy maximization principle and the geometric mean inequality are used in [9,17,16] and [14] respectively. An objective function that consists of three terms: the size of overlapping area, the camera motion parameters, and the mean of registration errors of tentative correspondences is optimized in [21] for free form shape registration.…”

Section: Previous Workmentioning

confidence: 99%