Orientation Estimation in 3D for Steerable Filters
Anshuman Sinha
Our project is to extend a previous work in the lab: 'Principled Design and Implementation of Steerable Detectors' from 2D to 3D. The complete pipeline will involve the detection of patterns of interest in a 3D image - both its position and orientation in space.
My contribution addresses the specific task of determining the unknown orientation of a known 3D template. Generalising existing 2D tools, especially for the modelling and accurate estimation of rotations in 3D, poses a certain degree of complexity. I will present a gradient-based stochastic method to tackle this problem. With a brief analysis on the results of the test experiments conducted, we hope to initiate a discussion on how this part of the project reflects on the scope of the larger one, and beyond.
Anshuman Sinha
Meeting • 2023-04-18
AbstractDeveloping robust frameworks for template detection in 3D imaging is of practical significance across multiple disciplines.Our project is to extend a previous work in the lab: 'Principled Design and Implementation of Steerable Detectors' from 2D to 3D. The complete pipeline will involve the detection of patterns of interest in a 3D image - both its position and orientation in space.
My contribution addresses the specific task of determining the unknown orientation of a known 3D template. Generalising existing 2D tools, especially for the modelling and accurate estimation of rotations in 3D, poses a certain degree of complexity. I will present a gradient-based stochastic method to tackle this problem. With a brief analysis on the results of the test experiments conducted, we hope to initiate a discussion on how this part of the project reflects on the scope of the larger one, and beyond.