DOI: 10.14489/vkit.2014.02.pp.016-021

Тихомирова Т.А., Федоренко Г. Т., Кириллова Л. Н.  
ФЛИС-ТРАССИРОВКА ЗАМКНУТЫХ И ВЕТВЯЩИХСЯ КОНТУРОВ
(с. 16–21)

Аннотация. Представлен алгоритм быстрой флис-трассировки, дополненный новыми типами связности контуров – «розетка» и «дендрит». Показано, что такое дополнение существенно повышает надежность считывания ветвящихся контуров, а также абрисов объектов, периметр которых мал по сравнению с длиной трассы.

Ключевые слова: контурное изображение; быстрая трассировка; распознавание образов.

Tikhomirova T. A., Fedorenko G. T., Kirillova L. N.
 FLEECE-TRACING OF CLOSED AND BRANCHED CURVES
(pp. 16–21)

Abstract. The contour objects are not always present in the image explicitly. However, many types of information can be converted into visible contours. This transformation is called segmentation. The type of information being converted into contours depends on the conditions of observation. Contour objects associated with visible boundaries, we call the outlines. They are characterized by great length and lack of branching. Outlines are the most useful for navigation, but their shapes in medium scale rarely are unique. Therefore, recognition of outlines may require additional information. If there are no clear boundaries or their segmentation does not give clear contours we should visualize centerlines of texture objects. Such centerlines we call the dendrites, as it is their typical shape. Usually dendrites are rather small, so they can't be used as a stand-alone landmark. However, dendrites are more unique than the outlines, and can effectively complement the latter. Dendrites are visualized by broadening the boundaries of texture objects. In result the size the central regions is reduced, and their contrast increases. Usually, the dendrites have a smoother shape compared to the outlines. Therefore, this procedure can be used to smooth outlines, if their shape is too complicated or unreliable. Examples – edge of the forest, coastline, etc. In this paper we present fast algorithm for dendrite fleece-tracing. The basic fleece-tracing algorithm provides simple tracks where all fragments are connected sequentially (“outlines” connectivity). In this case branched curves are presented as a number of independent tracks, and recognition of such object is complicated. At the same time, closed curves been produced by small-size isolated objects could be missed if tracks are long. In this paper we present a fleece-tracing algorithm with additional types of track's connectivity – “socket” and “dendrite”. It is shown that new types of connectivity significantly increases the reliability of detection of the branched curves and small objects. Moreover, in this case the branched curve is presented by single track, so it can be recognized as a single object.

Keywords: Contour images; Fast tracing; Pattern recognition.