Allocation of three brightness levels on a noisy image
Keywords:image reconstruction, verification of statistical hypotheses, binary classification
A new recovery method for images with three unknown brightness levels is proposed. In order to determine these levels, we use the image fragments whose histograms correspond to a given noise distribution. All pixels are distributed over the found brightness levels by a binary classification. The numerical results show the error in the estimate of the original brightnesses is no more than 3%. When the noise level is relatively low, the fraction of wrong classified pixels in their total amount is less than 0.006.
- A. A. Potapov, A. A. Pakhomov, S. A. Nikitov, and Yu. V. Gulyaev, Novel Methods of Image Processing (Fizmatlit, Moscow, 2008) [in Russian].
- A. R. Dabagov, I. A. Malyutina, and S. A. Filist, Artificial Intelligence Systems for X-Ray Examinations in Digital Medicine (Universitetskaya Kniga, Kursk, 2019).
- A. M. Golubkov, “Face Recognition Using Images Binary Classification Methods,” Izv. Saint Petersburg Eletrotekh. Univ., No. 7, 26-30 (2018).
- A. Rumyantsev, R. Minyazev, S. Dyganov, et al., “Assessment of the Influence of the Neural Network Architecture Size on the Training Rate in the Problem of Binary Classification,” Vestn. Tekhnol. Univ. 21 (8), 124-127 (2018).
- S. A. Aivazian, V. M. Buchstaber, I. S. Yenyukov, and L. D. Meshalkin, Applied Statistics. Vol. 3: Classification and Reduction of Dimensionality (Financy i Statistika, Moscow, 1989) [in Russian].
- I. S. Gruzman, V. S. Kirichuk, V. P. Kosykh, et al., Digital Processing of Images in Information Systems (Novosibirsk Gos. Tekh. Univ., Novosibirsk, 2002) [in Russian].
- V. S. Pugachev, Theory of Probability and Mathematical Statistics (Nauka, Moscow, 2002) [in Russian].
- M. E. Karasikov and Y. V. Maximov, “Dimensionality Reduction for Multi-Class Learning Problems Reduced to Multiple Binary Problems,” Mashinnoe Obuchenie Analiz Dannykh 1 (9), 1273-1290 (2014).
- A. V. Likhachov, “Tomographic Reconstruction of a Region with a Given Density Value,” Vychisl. Metody Programm. 19, 516-521 (2018).
- A. V. Likhachev, “Modified Method for Detecting Small Structures in Noisy Images,” Avtometriya 55 (6), 55-63 (2019) [Optoelectr., Instrum. Data Process. 55 (6), 580-586 (2019)].