Full orientation invariance and improved feature selectivity of 3D SIFT with application to medical image analysis

Allaire, Stéphane; Kim, J.J.; Breen, Stephen; Jaffray, David A.; Pekar, Vladimir

doi:10.1109/cvprw.2008.4563023

Cited by 93 publications

(108 citation statements)

References 12 publications

Supporting

Mentioning

101

Contrasting

Unclassified

Order By: Relevance

“…This descriptor is closely modelled on that used in [6,50], considering the correct definition of 3D orientation, based on azimuth, elevation and tilt. Volume gradients are examined in a two stage process to locally establish an invariant orientation to be used in the subsequent keypoint description.…”

Section: D Scale-invariant Feature Transform (Sift)mentioning

confidence: 99%

Object classification in 3D baggage security computed tomography imagery using visual codebooks

Flitton

Mouton

Breckon

2015

Pattern Recognition

View full text Add to dashboard Cite

Use policyThe full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for personal research or study, educational, or not-for-pro t purposes provided that:• a full bibliographic reference is made to the original source • a link is made to the metadata record in DRO • the full-text is not changed in any way The full-text must not be sold in any format or medium without the formal permission of the copyright holders.Please consult the full DRO policy for further details. AbstractWe investigate the performance of a Bag of (Visual) Words ( Optimal performance is achieved using the DH and DGH descriptors in conjunction with an uncertainty assignment methodology. Successful detection rates in excess of 97% for handguns and 89% for bottles and false-positive rates of approximately 2-3% are achieved. We demonstrate that the underlying imaging modality and the irrelevance of illumination and scale invariance within the transmission imagery context considered here, result in the favourable performance of simpler density histogram descriptors (DH, DGH) over 3D extensions of the well-established SIFT and RIFT feature descriptor approaches.

show abstract

Section: D Scale-invariant Feature Transform (Sift)mentioning

confidence: 99%

Object classification in 3D baggage security computed tomography imagery using visual codebooks

Flitton

Mouton

Breckon

2015

Pattern Recognition

View full text Add to dashboard Cite

show abstract

“…In contrast to the well known 3-D SIFT developed by Scovanner et al [42], which added the time as an extra dimension to the original 2-D descriptors for classifying actions in video sequences (i.e. 3-D here refers to x , y and t ), the method proposed by Allaire et al [43] extended the descriptors from 2-D images (i.e. x and y ) to 3-D spaces (i.e.…”

Section: Tracking Methods and Data Post-processing! 1) Descriptionmentioning

confidence: 99%

Kinect-based Universal Range Sensor and its Application in Educational Laboratories

Zhang

Chang

et al. 2015

Int. J. Onl. Eng.

View full text Add to dashboard Cite

Abstract-Traditional data acquisition (DAQ) systems for obtaining 3-D range information consist of sensors, DAQ measurement hardware and a processor with software. This kind of DAQ system is adapted and calibrated for various applications, thus imposing significant up-front costs that hamper its broad usage in educational laboratories. The low-cost Microsoft Kinect has become a promising alternative solution that has already been widely adopted by consumers and offers a wide variety of opportunities for being used in many other areas. The Kinect's cameras are capable of producing high quality synchronized video that consists of both color and depth data. This enables the Kinect to compete with other sophisticated 3-D sensor DAQ systems in terms of performance criteria such as accuracy, stability, reliability and error rates. One of the most noticeable Kinect-based applications is the skeleton-based tracking of humans, which is made possible by its built-in human skeleton recognition functions. Other common usages of the Kinect are 3-D surface/scene reconstruction and object classification/recognition. However, Kinect-based developments that focus on the tracking of arbitrary objects have rarely been reported, mainly due to a lack of mature algorithms.In the first part of this paper, a three-stage approach for capturing general motions of objects will be introduced. This approach consists of point cloud pre-processing with a focus on computational efficiency, object tracking employing recognition and post-processing including motion analysis. This approach can be tailored to special cases, namely the algorithms focus more on computation efficiency when the objects of interest have simple shapes or colors, and they focus more on reliability for objects with complex geometries or textures. The second part of this paper describes the integration of the proposed DAQ system into a multi-player game-based laboratory environment. In this implementation, a physical experiment is triggered by the game avatars and the experimental data acquired by the Kinect and analyzed by the proposed algorithms are then fed back into the game environment and used to animate the experimental device.

show abstract

“…The Laplacian of Gaussian gets a high response not only on round structures but also on elongated structures such as edges and ridges. As in [4], [1], we remove the unstable key points caused by elongated structures and we reject the interest points in the background with an Otsu thresholding.…”

Section: Interest Points and Featuresmentioning

confidence: 99%

Image Comparison on the Base of a Combinatorial Matching Algorithm

Drayer

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. In this paper we compare images based on the constellation of their interest points. The fundamental technique for this comparison is our matching algorithm, that is capable to model miss-and multimatches, while enforcing one-to-one matches. We associate an energy function for the possible matchings. In order to find the matching with the lowest energy, we reformulate this energy function as Markov Random Field and determine the matching with the lowest energy by an efficient minimization strategy. In the experiments, we compare our algorithm against the normalized cross correlation and a naive forth-andback best neighbor match algorithm. 1

show abstract

Full orientation invariance and improved feature selectivity of 3D SIFT with application to medical image analysis

Abstract: This paper presents a comprehensive extension of the

Cited by 93 publications

References 12 publications

Object classification in 3D baggage security computed tomography imagery using visual codebooks

Object classification in 3D baggage security computed tomography imagery using visual codebooks

Kinect-based Universal Range Sensor and its Application in Educational Laboratories

Image Comparison on the Base of a Combinatorial Matching Algorithm

Contact Info

Product

Resources

About