INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES
The subject matter of the article are the methods of local spatial post-processing of images obtained as a result of statistical per-pixel classification of multichannel satellite images distorted by additive Gaussian noise. The aim is to investigate the effectiveness of some variants of post-classi...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
National Aerospace University «Kharkiv Aviation Institute»
2019-06-01
|
| Series: | Радіоелектронні і комп'ютерні системи |
| Subjects: | |
| Online Access: | http://nti.khai.edu/ojs/index.php/reks/article/view/759 |
| _version_ | 1797707649678049280 |
|---|---|
| author | Ирина Карловна Васильева Владимир Васильевич Лукин |
| author_facet | Ирина Карловна Васильева Владимир Васильевич Лукин |
| author_sort | Ирина Карловна Васильева |
| collection | DOAJ |
| description | The subject matter of the article are the methods of local spatial post-processing of images obtained as a result of statistical per-pixel classification of multichannel satellite images distorted by additive Gaussian noise. The aim is to investigate the effectiveness of some variants of post-classification image processing methods over a wide range of signal-to-noise ratio; as a criterion of effectiveness, observed objects classification reliability indicators have been taken. The tasks to be solved are: to generate random values of the noise components brightness, ensuring that they coincide with the adopted probabilistic model; to implement a procedure of statistical controlled classification by the maximum likelihood method for images distorted by noise; to evaluate the results of the objects selection in noisy images by the criterion of the empirical probability of correct recognition; to implement procedures for local object-oriented post-processing of images; to investigate the effect of noise variance on the effectiveness of post-processing procedures. The methods used are: methods of stochastic simulation, methods of approximation of empirical dependencies, statistical methods of recognition, methods of probability theory and mathematical statistics, methods of local spatial filtering. The following results have been obtained. Algorithms of rank and weighted median post-processing with considering the results of classification by k-nearest neighbors in the filter window were implemented. The developed algorithms efficiency analysis that based on estimates of the correct recognition probability for objects on noisy images was carried out. Empirical dependences of the estimates of the overall recognition errors probability versus the additive noise variance were obtained. Conclusions. The scientific novelty of the results obtained is as follows: combined approaches to building decision rules, taking into account destabilizing factors, have been further developed – it has been shown that the use of methods of local object-oriented filtering of segmented images reduces the number of point errors in the element-based classification of objects, as well as partially restores the connectedness and spatial distribution of image structure elements. |
| first_indexed | 2024-03-12T06:10:30Z |
| format | Article |
| id | doaj.art-3e19b34eb6704d5988f0a3370fb374ca |
| institution | Directory Open Access Journal |
| issn | 1814-4225 2663-2012 |
| language | English |
| last_indexed | 2024-03-12T06:10:30Z |
| publishDate | 2019-06-01 |
| publisher | National Aerospace University «Kharkiv Aviation Institute» |
| record_format | Article |
| series | Радіоелектронні і комп'ютерні системи |
| spelling | doaj.art-3e19b34eb6704d5988f0a3370fb374ca2023-09-03T03:04:29ZengNational Aerospace University «Kharkiv Aviation Institute»Радіоелектронні і комп'ютерні системи1814-42252663-20122019-06-0102455910.32620/reks.2019.2.04802INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGESИрина Карловна Васильева0Владимир Васильевич Лукин1Национальный аэрокосмический университет им. Н. Е. Жуковского «Харьковский авиационный институт», ХарьковНациональный аэрокосмический университет им. Н. Е. Жуковского «Харьковский авиационный институт», ХарьковThe subject matter of the article are the methods of local spatial post-processing of images obtained as a result of statistical per-pixel classification of multichannel satellite images distorted by additive Gaussian noise. The aim is to investigate the effectiveness of some variants of post-classification image processing methods over a wide range of signal-to-noise ratio; as a criterion of effectiveness, observed objects classification reliability indicators have been taken. The tasks to be solved are: to generate random values of the noise components brightness, ensuring that they coincide with the adopted probabilistic model; to implement a procedure of statistical controlled classification by the maximum likelihood method for images distorted by noise; to evaluate the results of the objects selection in noisy images by the criterion of the empirical probability of correct recognition; to implement procedures for local object-oriented post-processing of images; to investigate the effect of noise variance on the effectiveness of post-processing procedures. The methods used are: methods of stochastic simulation, methods of approximation of empirical dependencies, statistical methods of recognition, methods of probability theory and mathematical statistics, methods of local spatial filtering. The following results have been obtained. Algorithms of rank and weighted median post-processing with considering the results of classification by k-nearest neighbors in the filter window were implemented. The developed algorithms efficiency analysis that based on estimates of the correct recognition probability for objects on noisy images was carried out. Empirical dependences of the estimates of the overall recognition errors probability versus the additive noise variance were obtained. Conclusions. The scientific novelty of the results obtained is as follows: combined approaches to building decision rules, taking into account destabilizing factors, have been further developed – it has been shown that the use of methods of local object-oriented filtering of segmented images reduces the number of point errors in the element-based classification of objects, as well as partially restores the connectedness and spatial distribution of image structure elements.http://nti.khai.edu/ojs/index.php/reks/article/view/759классификацияаппроксимацияаддитивный гауссов шумпространственная фильтрациявероятность правильного распознавания |
| spellingShingle | Ирина Карловна Васильева Владимир Васильевич Лукин INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES Радіоелектронні і комп'ютерні системи классификация аппроксимация аддитивный гауссов шум пространственная фильтрация вероятность правильного распознавания |
| title | INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES |
| title_full | INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES |
| title_fullStr | INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES |
| title_full_unstemmed | INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES |
| title_short | INVESTIGATION OF THE EFFICIENCY OF THE POST-CLASSIFICATION TECHNIQUES FOR NOISY MULTI-CHANNEL IMAGES |
| title_sort | investigation of the efficiency of the post classification techniques for noisy multi channel images |
| topic | классификация аппроксимация аддитивный гауссов шум пространственная фильтрация вероятность правильного распознавания |
| url | http://nti.khai.edu/ojs/index.php/reks/article/view/759 |
| work_keys_str_mv | AT irinakarlovnavasilʹeva investigationoftheefficiencyofthepostclassificationtechniquesfornoisymultichannelimages AT vladimirvasilʹevičlukin investigationoftheefficiencyofthepostclassificationtechniquesfornoisymultichannelimages |