RADIOELECTRONIC AND COMPUTER SYSTEMS

The subject of the article is to analyze the effectiveness of a new method for detecting heterogeneities in a digital image by estimating the bimagnitude maximum of the pixel intensities. The aim is to evaluate the effectiveness of the new method of detecting heterogeneities in the image using the maximum of the bimagnitude compared to the known method based on the estimation of the local root mean square deviation (LRMSD) of pixel intensity values. The objectives of the paper are the following: to formalize the procedure for computing the bimagnitude maximum of the pixels in the local segment; create a test image with different contrast values on the borders; to develop a mathematical model for calculating in the Matlab system the efficiency of detecting heterogeneities in the image in the presence of additive Gaussian noise with different values of noise RMS; provide for analysis and comparison of the graphs the receiver operating characteristic (ROC) contained the number of correctly classified non-homogeneous areas versus the number of incorrectly classified areas. The used methods are the following: bispectral data analysis method; methods of probability theory and mathematical statistics; methods of digital image processing. The following results were obtained. A boundary map for the test image without distortion and the presence of additive Gaussian noise with a variance equal to 0.2 is constructed for two types of detectors: the first one is based on the maximum amplitude and the second one is based on the estimation of the local RMS. The results of computer simulations show that both detectors fine-tune the boundary for the images in the absence of noise. But in the presence of additive noise, the detector based on the biamplitude maximum provides a significant advantage. Graphs of the dependence of the number of correctly classified inhomogeneous sections on the number of incorrectly classified areas for the proposed and known reference detection methods are represented. The area under the curve (AUC) values that characterize the efficiency of detecting heterogeneities in the image are calculated. The scientific novelty of the obtained results is the following: a new approach of detecting inhomogeneities in the image is proposed with the help of a new informative feature estimated in the form of the local biamplitude maximum. To analyze the effectiveness of the proposed method, a test image was formed with different border contrast values. Using the proposed technique and the known method, boundary maps were constructed for the test image without distortion and in the presence of additive Gaussian noise. To evaluate the effectiveness of two methods, the graphs were plotted against the number of correctly classified inhomogeneous sites by the number of incorrectly classified (ROC) for both proposed and known detection methods. A detector based on the local RMS value is more effective at small Gaussian noise variance values, but as the noise variance increases, detector based on the biamplitude maximum estimation is more effective. The calculated AUC values for studied methods based on local RMS estimation and maximum biamplitude estimation are equal to 0. 678 and 0. 8468, respectively. Even though the proposed method loses efficiency, the bispectrum-based method is more effective at large values of noise variance, in particular, when the noise RMS is 0.6, AUC = 0.8748.

bispectrum; biamplitude; image processing; ROC curve; AUC parameter

Erinjery, J. J., Singh, M., Kent, R. Mapping and assessment of vegetation types in the tropical rainforests of the Western Ghats using multispectral Sentinel-2 and SAR Sentinel-1 satellite imagery. Remote Sens. Environment, October 2018, vol. 216, pp. 345-354.

Tollerud, H. J., Brown, J. F., Loveland, T. R. Investigating the Effects of Land Use and Land Cover on the Relationship between Moisture and Reflectance Using Landsat Time Series. Remote Sensing, June 2020, vol. 12, pp. 1-29.

Haralick, R. M., Shanmugam, K., Dinstein, I. Textural features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, vol. 3, no. 6, 1973, pp. 610-621.

Terraien, Ch. U., Kuat'eri, T. F., Dadzhon, D. E. Algoritmy analiza izobrazhenii, osnovannye na statisticheskikh modelyakh [Statistical Model Image Analysis Algorithms]. vol. 74, no. 4, 1986, pp. 4-25.

Basvil, M., Vilsky, A., Banvenist, A. Obnaruzhenie izmeneniya svoistv signalov i dinamicheskikh sistem [Detection of changes in the properties of signals and dynamic systems]. Moscow, Mir Publ., 1989. 278 p. ISBN 5-03-000573-0.

Malakhov, A. N. Kumulyantnyi analiz sluchai-nykh negaussovykh protsessov i ikh preobrazovanii [Cumulative analysis of random non-Gaussian processes and their transformations]. Moscow, Soviet radio Publ., 1978. 376 p.

Nikias, C. L., Raughuveer, M. R. Bispectral estimation: A digital signal processing framework. Proc. IEEE, vol. 75, no. 7, 1987, pp. 869-891.

Astola, J. T., Egiazarian, K. O., Khlopov, G. I., Khomenko, S. I., Kurbatov, I. V., Ye. Morozov, V., Totsky, A. V. Application of bispectrum estimation for time-frequency analysis of ground surveillance Doppler radar echo signals. IEEE Transactions on Instrumentation and Measurement, vol. 57, no. 9, 2008, pp. 1949-1957.

Ling, X. C., Huang, J., Zhang, H. AUC: a Statistically Consistent and more Discriminating Measure than Accuracy. IJCAI'03: Proceedings of the 18th international joint conference on Artificial intelligence, 2003, pp. 519-524.

Ling, X. C., Huang, J., Zhang, H. AUC: a Better Measure than Accuracy in Comparing Learning Algorithms. Advances in Artificial Intelligence: 16th Conference of the Canadian Society for Computational Studies of Intelligence, 2003, pp. 329-341.

Belikova, T. P., Yaroslavskij, L. P. Ispolzovanie adaptivnyh amplitudnyh preobrazovanij dlya preparirovaniya izobrazhenij [Using adaptive amplitude transforms for image preparation]. Voprosy radioelektroniki. Ser. Obshetehn., 1974, vol. 14, pp. 88-98.

Roenko, A. A., Fevralev, D. V., Ponomarenko, N. N., Lukin, V. V. Primenenie ustoichivykh otsenok parametrov vyborok dannykh pri obrabotke izobrazhenii [Application of stable estimates of parameters of data samples in image processing]. Vostochnoevropeiskii zhurnal peredovykh tekhnologii, 2007, no. 3/2 (27), pp. 21-31.

Melnik, V. P., Lukin, V. V., Zelensky, A. A., Astola, J. T. Kuosmanen, P. Local activity indicators: analysis and application to hard-switching adaptive filtering of images. Optical Engineering Journal, 2001, vol. 40, no. 8, pp. 1441-1445.