RESEARCH OF IMAGE THEORY: PREPROCECION – EDGE DETECTORS
DOI:
https://doi.org/10.32782/mathematical-modelling/2023-6-1-2Keywords:
evaluation, edge detectors, computer visionAbstract
For decades, pattern recognition and image processing, in particular, has remained an urgent task. Today we have a well-developed theoretical foundation of basic concepts, operations and the need to comparison of basic refining operations of the input (initial) information, which allows the effective application of the theory of recognition. In the paper, the authors present a partial solution, having investigated the possibilities of ten operators for selecting edges (contours). The work compares the behavior of operators in different conditions. In particular, the conditions are different values of the ranges of the visible spectrum. The visible spectrum is divided into four bands, which are effectively perceived by the light-sensitive matrix of the webcam. These ranges correspond to different spectrums of daylight. The following operators were investigated in the work: Edge Enhancement by Discrete Differentiating, Wallis logarithmic edge detection, Frei-Chen edge and line detection, edge detector Kirsch Edge detector, Directional edge detection, Product of the difference of averages, Crack Edge detection, Marr-Hildreth edge detection, local edge detection in three-dimensional images (Local Edge Detection in three-dimensional images), hierarchical edge detection (Hierarchical edge detection). To conduct research, a system of adaptive search for the optimal image edge detector according to the level of illumination has been developed. This system is designed to work with a web camera of a personal computer or smartphone, selecting objects on it and then applying filters to them. As a result, the system will select the best method that has been determined for certain lighting. Methods and results are presented in the paper. The work covers the development and application of edge extraction methods to detect significant features and features in images using an adaptive approach as opposed to a fixed sequence of pre-processing. The tasks were solved during the research: the methods of identifying the key characteristics of the object were implemented; implemented a part of the analytical core with an objective estimate that is better than MSE and PSNR and their nearest variations. In particular, the structural similarity index (SSIM) was used. A part of the analytical core for making a decision based on the results of an objective evaluation was implemented, which allowed adaptively determining the optimal algorithm for finding objects in the field of attention, a database of reference vectors of human perception of objects was created, and an educational component was developed for training computer vision with human concepts images of numbers.
References
Hrytsyk V., Grondzal A., Bilenkyj A. Augmented reality for people with disabilities. CSIT’2015. Lviv : LPNU, 2015. P. 188–191.
Hrytsyk V., Nazarkevych M. Real-Time Sensing, Reasoning and Adaptation for Computer Vision Systems. Lecture Notes on Data Engineering and Communications Technologies. 2022. Vol. 77. Р. 573–585.
Cutting-edge facial recognition goes mainstream. Reasearch*eu results magazine. December 2017 – January 2018. Vol. 68. Р. 39.
Hrytsyk V., Medykovskyy M., Nazarkevych M. Estimation of Symmetry in the Recognition System with Adaptive Application of Filters. Symmetry. 2022. Vol. 14. № 5.
Competitive Diffusion of Gases in a Zeolite Bed: NMR and Slice Selection Procedure, Modeling, and Parameter Identification / M. Petryk et al. Journal of Physical Chemistry C. 2015. Vol. 119. № 47. Р. 519–525.
Krak I., Barmak O., Manziuk E. Using visual analytics to develop human and machine-centric models: A review of approaches and proposed information technology. Computitional Intelligence. 2020. № 26. P. 20–26.
Szeliski R. Computer Vision. London : Springer, 2011. Р. XX.
Katsaggelos A.K. Fundamentals of Digital Image and Video Processing. Coursera. 2020. URL: https://www.coursera.org/learn/digital
Pitsun О. OpenCV image segmentation. 2016. URL: https://github.com/olehpitsun/CVforStudents/tree/master/src/task2
Scikit-learn. Machine learning in Python. 2023. URL: https://scikit-learn.org/stable/
Jeffrey A. Klark (Alex) and contributors. Pillow (Python Imaging Library) 9.5.0 documentation. Pillow. 2023. URL: https://pillow.readthedocs.io/en/stable/
Tensor Flow. Machine learning models. Convolutional neural network. URL: https://www.tensorflow.org/
Forsyth D., J. Ponce. Computer Vision: A Modern Approach. 1 August 2002.
Method of detecting special points on biometric images based on new filtering methods / М. Nazarkevych et al. Workshop on Cybersecurity Providing in Information and Telecommunication Systems, CPITS 2021. Kyiv, 2021.
Evaluation of the effectiveness of different image skeletonization methods in biometric security systems / М. Nazarkevych et al. International Journal of Sensors, Wireless Communications and Control. 2021. Vol. 11. № 5. Р. 542–552.
