Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
ABSTRACT Near-regular texture (NRT), denoting deviations from otherwise symmetric wallpaper patterns, is commonly observable in the real world. Existing lattice detection algorithms capture the underlying lattice of an NRT pattern and all of its individual texels, facilitating an automated analysis of NRT. Many real world images, as in those of zebrafish larval histology arrays, depart significantly from regularity and challenge the current state of the art wallpaper group theory-based lattice detection methods. We propose an alternative 2D lattice detection algorithm that exploits translation and reflec- tion symmetries and specific imaging cues. By outperforming existing methods on histology array images, our algorithm leads us towards complete automation of high-throughput histological image processing while broadening the spectrum of NRT computation. Index Terms —lattice estimation, biological tissues, biomedical image processing 1. INTRODUCTION AND BACKGROUND Near-regular texture (NRT) is defined [1] as minor geometric, photometric, and topological deformations from an otherwise translationally symmetric 2D wallpaper pattern. NRT can be ubiquitously observed in both man-made and natural objects and their images, such as a checkerboard pattern on a wrinkled tablecloth, a brick wall with color and texture variations among the individual bricks, the hexagonal cells of natural honeycomb, or the scales that make up the curved surface of a fish or shark. Significant progress has been made in the automated detec- tion of NRT patterns, whether static [1, 2, 3] or in motion [4]. The detection algorithms used in these works are based on wallpaper group theory [5] in that the topology of such pat- terns can be characterized completely by a quadrilateral “lat- tice” composed of fundamental generating units, called texels or textons . Indeed, regular quadrilateral lattices can be repre- sented sufficiently by only two linearly independent vectors . B.C. (corresponding author – email: is with the Inte- grative Biosciences Program and is supported by the Penn State Academic Computing Fellowship. G.T. and K.C. are with the College of Medicine and are supported by NIH, the Life Sciences Greenhouse of Central Penn- sylvania, and Pennsylvania Department of Health Tobacco Settlement Funds. J.Z.W. is also with Carnegie Mellon Univ. (CMU) and is supported by NSF. Y.L. is also with CMU, as well as the University of Pittsburgh, and is supported by NSF, NIH, and the PA Department of Health. AUTOMATIC LATTICE DETECTION IN NEAR-REGULAR HISTOLOGY ARRAY IMAGES Brian A. Canada, Georgia K. Thomas, Keith C. Cheng, James Z. Wang, and Yanxi Liu The Pennsylvania State University, University Park, PA, USA Fig. 3. Output from our algorithm: The lattice overlaid on the rotation-corrected input array image is detected by exploiting the topological 2D lattice structure and maximizing bilateral symmetry of the zebrafish larva in each lattice unit cell. Fig. 1.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 4


This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online