Abstract:
Analysis of patterns in binary matrices plays a vital role in numerous applications of
computer science. One of the most essential patterns of such matrices are the so called
switching components, where the number and location of the components gives valuable
information about the binary matrix. One way to measure the effect of switching
components in a binary matrix is counting the number of 0-s which have to be replaced
with 1-s in order to eliminate the switching components. However, finding the minimal
number of 0-1 flips is generally an NP-complete problem. In the first part of the talk we
present two novel-type heuristics for the above problem and show via experiments that
they outperform the formerly proposed ones, both in optimality and in running time. [1]
In the second part of the talk we propose a new energy-minimization reconstruction model
for binary tomography. The model incorporates a shape orientation based regularization
term. Orientation of the object is used as a priori information in the reconstruction
process. Analytical and experimental analysis show, that the proposed method provides
reasonable results from very small amount of projection data, obtained from two or
particularly from just one projection direction. Therefore, the proposed method can be
valuable in situations of very limited projection availability. [2]
[1] N. Hantos, P. Balázs: Eliminating switching components in binary matrices by 0-1
flips and column permutations, Fundamenta Informaticae 141(2-3) 135-150 (2015)
[2] T. Lukic, P. Balázs: Binary tomography reconstruction based on shape orientation,
Pattern Recognition Letters, accepted (2016), doi:10.1016/j.patrec.2016.04.010
Two Recent Results on Binary Tomogr
Conferenza di Péter Balázs sala Tricerri del 27/05/2016 ore 11
23 Maggio 2016 (Archiviata)