Estimating the complexity of objects in images

Capa

Citar

Texto integral

Acesso aberto Acesso aberto
Acesso é fechado Acesso está concedido
Acesso é fechado Somente assinantes

Resumo

A new method for estimating the complexity of geometric shapes (spots) is proposed, taking into account the internal structure of the spots, and not only their external contour. The task of calculating the degree of complexity of objects is divided into components: segmentation of spots and estimation of the complexity of isolated spots. The new method has a relatively low computational complexity compared to the alternative methods considered in the work. Using the new method, an algorithm based on parallel computing of the CUDA programming language for graphics accelerators (video cards) was created, which further increases the performance of our method. A qualitative and quantitative analysis of existing (alternative) methods has been carried out, their advantages and disadvantages in comparison with our method and with each other have been revealed. The algorithm implemented on the basis of the new method has been tested on both artificial and real digital images.

Texto integral

Acesso é fechado

Sobre autores

V. Bokshanskiy

Bauman Moscow State Technical University

Email: shatskiyalex@gmail.com
Rússia, Moscow

V. Kulin

Bauman Moscow State Technical University

Email: shatskiyalex@gmail.com
Rússia, Moscow

G. Finiakin

Bauman Moscow State Technical University; National Research University “Moscow Power Engineering Institute”

Email: shatskiyalex@gmail.com
Rússia, Moscow; Moscow

A. Kharlamov

Moscow State Technical University of Civil Aviation

Email: shatskiyalex@gmail.com
Rússia, Moscow

A. Shatskiy

Bauman Moscow State Technical University

Autor responsável pela correspondência
Email: shatskiyalex@gmail.com
Rússia, Moscow

Bibliografia

  1. https://ru.wikipedia.org/wiki/WMAP/
  2. https://en.wikipedia.org/wiki/Planck_(spacecraft)
  3. Hu M.K. Visual pattern recognition by moment invariants, IRE Trans. Info. Theory, 1962. V. IT-8, P. 179–187 (1962).
  4. Doerr F.J.S., Florence, A.J. A micro-xrt image analysis and machine learning methodology for the characterisation of multi-particulate capsule formulations, Int. J. Pharm., 2020. V. 2.
  5. Kornilov A.S., Safonov I.V. An overview of watershed algorithm implementations in open source libraries, J. Imaging, 2018. V. 4. No. 10. P. 123.
  6. Shaiduk A.M., Ostanin S.A. Quantitative estimation of shape comlexity in medical images // Zh. Radioelektron. 2013. V. 2.
  7. Rothganger M., Melnik A., Ritter H. Shape complexity estimation using VAE. ArXiv:2304.02766, 2023.
  8. Gilchrist J. Parallel data compression with bzip2, Proceedings of the 16th IASTED International Conference on Parallel and Distributed Computing and Systems, 2004. V. 16. P. 559–564.
  9. Shannon C.E. A mathematical theory of communication. Bell Syst. Techn. J., 1948. V 27, No. 3. P. 379–423.

Arquivos suplementares

Arquivos suplementares
Ação
1. JATS XML
Baixar
2. Fig. 1. Image brightness intensity.

Baixar (140KB)
Metadados
3. Fig. 2. Grayscale images (left) and segmentation of the same images into individual segments (combined into spots) highlighted in different colors (right).

Baixar (305KB)
Metadados
4. Fig. 3. Sorting image spots by decreasing complexity level and highlighting the main spot with a green frame. The output of the top-5 spots of the image is shown, ranked first by the value of the Max_i number at the | kiHui | max value, and then by the spot brightness. On the right is a table of spot characteristics: spot number, Max_i number at the maximum value | kiHui | max, spot brightness Etot, spot area Stot and value | kiHui | max.

Baixar (263KB)
Metadados
5. Fig. 4. Scheme of the architecture of the neural network – variational autoencoder, example of tracing a test binary image and assessing its degree of complexity.

Baixar (229KB)
Metadados
6. Fig. 5. Test images used to compare complexity estimation methods for invariance to some of the affine transformations and to additive noise: apple (left), bird (center), and fly (right).

Baixar (43KB)
Metadados
7. Fig. 6. Stability under noise exposure obtained by different methods.

Baixar (99KB)
Metadados
8. Fig. 7. An example of a test image of a fly with noise implementation at RMS = 64(y. e.). For binary images, whose brightness values ​​are either 0 or 255 (at 8-bit coding depth), adding noise with RMS = 64 is not a factor, as a result of which the original geometric shapes cannot be visually recognized.

Baixar (134KB)
Metadados
9. Fig. 8. Change in relative difficulty for different methods when rotating test objects: apple (left), bird (center), and fly (right).

Baixar (344KB)
Metadados

Declaração de direitos autorais © Russian Academy of Sciences, 2024