Picture Segmentation Applications in Optometry and Vision Science

Jūs esate čia

Rekomenduojame aplankyti

Laboratorinė medicina. 2014,
t. 16,
Nr. 1,
p. 38 -
42

Summary

Introduction. Nowadays information about processes and different phenomena is presented via images, which is very useful because it is easiest way to present basic idea. Unfortunately presenting information in this way we often lose opportunity to look at data analytically. In this research we propose picture segmentation algorithm which is designed to identify regions within picture which share similar properties.

Material and methods. Proposed algorithm mimics ideas of split and merge algorithms with enhancements at multiple stages. At first stage of segmentation picture multithresholding algorithm is used in order to segment image judging by greyscale levels. Initial picture segmentation by colour is proposed as well. In following stages picture is divided into segments in way that ensures that all segment edge values are multiples of 2 regardless of initial image size and contains predefined portion of one colour or greyscale level pixels.

Results. Using proposed algorithm is possible to achieve reasonable distinction between objects and background. Algorithm is designed in way that all segmentation processes runs automatically and follow one another and doesn't require human intervention in process, still it is possible to change settings to save time or obtain more detailed segmentation.

Conclusion. We suggest that this algorithm could be useful in human machine interfaces, complex surface analysis in physics, as also vision science and graphics for data analysis and complex visual stimuli generation. Designing this algorithm we found slightly different approach of picture segmentation also identified new ideas which are worth developing.

© 2024, Lietuvos laboratorinės medicinos draugija