A Feature-Based Technique for Joint Linear Estimation of High-Order Image-to-Mosaic Transformations: Mosaicing the Curved Human Retina

  • Authors:
  • Ali Can;Charles V. Stewart;Badrinath Roysam;Howard L. Tanenbaum

  • Affiliations:
  • Rensselaer Polytechnic Institute, Troy, NY;Rensselear Polytechnic Institute, Troy, NY;Rensselear Polytechnic Institute, Troy, NY;The Center for Sight, Albany, NY

  • Venue:
  • IEEE Transactions on Pattern Analysis and Machine Intelligence
  • Year:
  • 2002

Quantified Score

Hi-index 0.14

Visualization

Abstract

An algorithm for constructing image mosaics from multiple, uncalibrated, weak-perspective views of the human retina is presented and analyzed. It builds on a previously described algorithm for registering pairs of retinal images using a noninvertible, 12-parameter, quadratic image transformation model and a hierarchical, robust estimation technique. The major innovation presented here is a linear, feature-based, noniterative method for jointly estimating consistent transformations of all images onto the mosaic 驴anchor image.驴 Constraints for this estimation are derived from pairwise registration both directly with the anchor image and indirectly between pairs of nonanchor images. An incremental, graph-based technique constructs the set of registered image pairs used in the joint solution. The joint estimation technique allows images that do not overlap the anchor frame to be successfully mosaiced, a particularly valuable capability for mosaicing images of the retinal periphery. Experimental analysis of the algorithm on data sets from 16 eyes shows the average overall median transformation error in final mosaic construction to be 0.76 pixels. Overall, the technique is simpler, more accurate, and offers broader coverage than previously published methods.