Advertisement
Published in:
2018 | OriginalPaper | Chapter
Vehicle X-Ray Images Registration
Authors : Abraham Marciano, Laurent D. Cohen, Najib Gadi
Published in: Energy Minimization Methods in Computer Vision and Pattern Recognition
Publisher: Springer International Publishing
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
Abstract
Image registration is definitely one of the most prominent techniques at the heart of computer vision research. Applications range from medical image analysis, remote sensing or robotics to security-related tasks such as surveillance or motion tracking. In our previous work, a solution was provided to address the registration problem involving top-view radiographic images of vehicles. A unidimensional minimization scheme was formulated along with a column-wise constancy constraint on the displacement field.
In this paper, we show that the proposed method is not sufficient in case of significant vertical shifts between the cars of both moving and static images. In fact, the radiated beam is triangular, thus any translated object is projected differently according to its distance to the x-ray source. We therefore add a 1D unconstrained optimization to the previous scheme for a y-direction correction. We also demonstrate that applying the vertical correction following our 1D optimization in the x-axis yields better results than performing a simultaneous minimization on both components.
Finally, the possible apparition of artefacts in the deformed image throughout the optimization process is analyzed. Diffusion and volume-preserving schemes are considered and compared in this regard.