Evrişimsel Sinir Ağları ile Görüntülerde Gürültü Türünü Saptama

Öz

Gürültü, görüntü elde etme sırasında görüntüye eklenen istenmeyen sinyallerdir. Bir görüntüden gürültünün arındırılmasında kullanılan filtre yöntemlerinin başarılı olabilmesi için gürültü türünün doğru şekilde analiz edilmesi gerekmektedir. Bu çalışma ile görüntülerdeki gürültü türünün ve gürültüsüz görüntülerin doğru ve pratik şekilde saptanması hedeflenmiştir. Ayrıca, Evrişimli Sinir Ağları (ESA) ile gürültü tahmininde hangi eniyileme algoritmasının tercih edilebileceğine ışık tutulmaya çalışılmıştır. Görüntülerde tuz-biber, gauss ve benek gürültü türlerinin saptanması için VGG-16 mimarisi temel alınarak bir ESA modeli önerilmiştir. Önerilen model transfer öğrenme yöntemi ve ince ayar yaklaşımı kullanılarak eğitilmiş ve beş eniyileme algoritmasının model başarımı üzerindeki etkisi incelenmiştir. Gürültü türünün saptanması için en iyi doğruluk %98,75 ile RMSProp eniyileme algoritması kullanılarak elde edilmiştir. Başarım performansları, gürültü türünün saptanmasında önerilen ESA mimarisinin başarı ile kullanılabileceği gösterilmiştir.

Anahtar Kelimeler

• Gürültü Türünü Saptama
• Transfer Öğrenme
• Eniyileme Algoritmaları

Noise Type Detection in Images with Convolutional Neural Networks

Öz

Noise is unwanted signals added to the image during image acquisition. In order for the filter methods used to remove noise from an image to be successful, the type of noise must be analyzed correctly. This study aims to identify the type of noise in images and noise-free images accurately and practically. In addition, it is tried to highlight on which optimization algorithm can be preferred in noise estimation with Convolutional Neural Networks (CNN). A CNN model based on the VGG-16 architecture was proposed for the detection of salt and pepper, Gaussian and speckle noise types in images. The proposed model was trained using the transfer learning method and fine-tuning approach, and the effect of five optimization algorithms on the model performance was investigated. The best accuracy of 98.75% for noise type detection was obtained using the RMSProp optimization algorithm. The performance results show that the proposed CNN architecture can be successfully used for noise type detection.

Anahtar Kelimeler

• Keywords Noise Type Detection
• Transfer Learning
• Optimization Algorithms

Kaynakça

1. Akar, E., Kara S., Akdemir H., ve Kırış A. A MATLAB tool for an easy application and comparison of image denoising methods. In 2015 Medical Technologies National Conference (TIPTEKNO), pp. 1-4. IEEE, 2015.
2. Hoomod, H. K., ve Dawood S. H. Fast image denoising based on modify CNN and noise estimation. 2017 Annual Conference on New Trends in Information & Communications Technology Applications (NTICT). IEEE, 2017.
3. Verma, R., ve Ali J. A comparative study of various types of image noise and efficient noise removal techniques. International Journal of advanced research in computer science and software engineering 3.10 (2013).
4. Değirmenci, Ali, Çankaya İ., ve Demirci R. Gradyan Anahtarlamalı Gauss Görüntü Filtresi. Düzce Üniversitesi Bilim ve Teknoloji Dergisi 6.1 (2018): 196-215.
5. Magud, O., Tuba E., ve Bacanin N.. An algorithm for medical ultrasound image enhancement by speckle noise reduction. International Journal of Signal Processing 1 (2016): 146-151.
6. Kaur, S. Noise types and various removal techniques. International Journal of Advanced Research in Electronics and Communication Engineering (IJARECE) 4.2 (2015): 226-230.
7. Dong, Shi, Wang P., ve Abbas K. A survey on deep learning and its applications. Computer Science Review 40 (2021): 100379.
8. Kaya, M. Feature fusion-based ensemble CNN learning optimization for automated detection of pediatric pneumonia. Biomedical Signal Processing and Control 87 (2024): 105472.

9. Akşehir, Z. D., ve Kılıç E. Hisse Senedi Tahmininde Karşılaşılan Veri Dengesizliği Problemi için Yeni Bir Kural Tabanlı Yaklaşım ve 2D-CNN Modeli. Türkiye Bilişim Vakfı Bilgisayar Bilimleri ve Mühendisliği Dergisi 15.1 (2022): 6-13.
10. Karn, A.L., Sengan, S., Kotecha, K., Pustokhina, I.V., Pustokhin, D.A., Subramaniyaswamy, V., ve Buddhi, D. ICACIA: An Intelligent Context-Aware framework for COBOT in defense industry using ontological and deep learning models. Robotics and Autonomous Systems 157 (2022): 104234.
11. Koklu, M., Unlersen F. M., Ozkan İ. A., Aslan M. F., ve Sabanci K. A CNN-SVM study based on selected deep features for grapevine leaves classification. Measurement 188 (2022): 110425.
12. Kaya, M., Ulutürk Samet, Çetin Kaya Y., Altıntaş O., ve Turan B. Optimization of Several Deep CNN Models for Waste Classification. Sakarya University Journal of Computer and Information Sciences 6, no. 2 (2023): 91-104.
13. Alzubaidi, L., Zhang, J., Humaidi, A.J., Al-Dujaili, A., Duan, Y., Al-Shamma, O., Santamaría, J., Fadhel, M.A., Al-Amidie, M., ve Farhan, L.,. Review of deep learning: Concepts, CNN architectures, challenges, applications, future directions. Journal of big Data 8 (2021): 1-74.
14. Reddy, S. V. G., Reddy K. T., ve ValliKumari V.. Optimization of deep learning using various optimizers, loss functions and dropout. Int. J. Recent Technol. Eng 7 (2018): 448-455.
15. Ward, R., Wu X., ve Bottou L. Adagrad stepsizes: Sharp convergence over nonconvex landscapes. The Journal of Machine Learning Research 21.1 (2020): 9047-9076.
16. Kingma, D. P., ve Adam J. B. A method for stochastic optimization. arXiv preprint arXiv:1412.6980 (2014).
17. Zeiler, M. D. Adadelta: an adaptive learning rate method. arXiv preprint arXiv:1212.5701 (2012).
18. Darken, C., Chang, J., ve Moody J. Learning rate schedules for faster stochastic gradient search. Neural networks for signal processing. Vol. 2. Helsinoger, Denmark: Citeseer, 1992.
19. Phan, T. V., Sultana, S., Nguyen, T. G., ve Bauschert, T. Q-TRANSFER: A Novel Framework for Efficient Deep Transfer Learning in Networking. In 2020 International Conference on Artificial Intelligence in Information and Communication (ICAIIC), pp. 146-151. IEEE, 2020.
20. Tai, S-C, ve Yang S-M. A fast method for image noise estimation using laplacian operator and adaptive edge detection. 2008 3rd International Symposium on Communications, Control and Signal Processing. IEEE, 2008.
21. Zoran, D., ve Weiss, Y. Scale invariance and noise in natural images. 2009 IEEE 12th International Conference on Computer Vision. IEEE, 2009.
22. Burger, H. C., Schuler, C. J., ve Harmeling. S. Image denoising: Can plain neural networks compete with BM3D?. 2012 IEEE conference on computer vision and pattern recognition. IEEE, 2012.
23. Pimpalkhute, V. A., Page R., Kothari A., Bhurchandi, K. M., ve Kamble, V. M. Digital image noise estimation using DWT coefficients. IEEE transactions on image processing 30 (2021): 1962-1972.
24. Güraksın, G. E. Tuz Biber Gürültülerinin Giderilmesi için k-Ortalama Algoritması Tabanlı Filtre Tasarımı. Süleyman Demirel Üniversitesi Fen Bilimleri Enstitüsü Dergisi, 2018, 22 (2), 972-978.
25. Küpeli, C., ve Bulut, F. Görüntüdeki Tuz Biber ve Gauss Gürültülerine Karşı Filtrelerin Başarım Analizleri. Haliç Üniversitesi Fen Bilimleri Dergisi 3.2 (2020): 211-239.
26. Sil, D., Dutta, A., ve Chandra, A. Convolutional neural networks for noise classification and denoising of images. In TENCON 2019-2019 IEEE Region 10 Conference (TENCON), pp. 447-451. IEEE, 2019.
27. Liu, F., Song, Q., ve Jin, G. The classification and denoising of image noise based on deep neural networks. Applied Intelligence 50 (2020): 2194-2207.
28. Lemarchand, F., Findeli, T., Nogues, E., ve Pelcat. M. Noisebreaker: Gradual image denoising guided by noise analysis. In 2020 IEEE 22nd International Workshop on Multimedia Signal Processing (MMSP), pp. 1-6. IEEE, 2020.
29. Li, Y., Yu X., Pei, J., Huo, W., Zhang, Y., Huang, Y., ve Yang, J.. A Learning-Based Multi-Type Noise Suppressing Method for Remote Sensing Images. In IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Symposium, pp. 3299-3302. IEEE, 2022.
30. Tripathy, A., Das, A., Patel, M., Singhai, E., ve Tripathy, S. Transfer Learning based Noise Classification in Chest X-Ray Images. In 2021 Smart Technologies, Communication and Robotics (STCR), pp. 1-7. IEEE, 2021.
31. Rahman, S. S. M. M., Salomon, M., ve Dembélé, S. Machine learning aided classification of noise distribution in scanning electron microscopy images. In 2023 3rd International Conference on Computer, Control and Robotics (ICCCR), pp. 111-115. IEEE, 2023.
32. Simonyan, K., ve Zisserman, A. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556 (2014).
33. Tao, J., Gu, Y., Sun, J.Z., Bie, Y., ve Wang, H. Research on vgg16 convolutional neural network feature classification algorithm based on Transfer Learning. In 2021 2nd China international SAR symposium (CISS), pp. 1-3. IEEE, 2021.
34. Fan, J., Lee J. H., ve Lee YK. Application of transfer learning for image classification on dataset with not mutually exclusive classes. 2021 36th International Technical Conference on Circuits/Systems, Computers and Communications (ITC-CSCC). IEEE, 2021.
35. Fırıldak, K., ve Talu, M. F. Evrişimsel sinir ağlarında kullanılan transfer öğrenme yaklaşımlarının incelenmesi. Computer Science 4.2 (2019): 88-95.
36. Rajeswari, S. S., ve Nair, M. A Transfer Learning Approach for Predicting Alzheimer's Disease. 2021 4th Biennial International Conference on Nascent Technologies in Engineering (ICNTE). IEEE, 2021.
37. Li, G., Zhen, H., Jiao,F., Hao, T., Wang, D., ve Ni, K. Research on tobacco leaf grading algorithm based on transfer learning. In 2021 IEEE International Conference on Artificial Intelligence and Computer Applications (ICAICA), pp. 32-35. IEEE, 2021.
38. Seyyarer, E., Ayata, F., Uçkan, T., ve Karci, A. Derin öğrenmede kullanilan optimizasyon algoritmalarinin uygulanmasi ve kiyaslanmasi. Computer Science 5, no. 2 (2020): 90-98.
39. Quinn, J., McEachen, J., Fullan, M., Gardner, M., ve Drummy, M. Dive into deep learning: Tools for engagement. Corwin Press, 2019.
40. Goodfellow, I., Bengio, Y., ve Courville, A. Deep learning. MIT press, 2016.
41. Bisong, E. Building machine learning and deep learning models on Google cloud platform. Berkeley, CA: Apress, 2019.
42. Géron, A. Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow. O'Reilly Media, Inc.", 2022.
43. Madhavan, S., Ahmed, S., Rao, V., ve John, A., 2021. Compare deep learning frameworks.
44. https://developer.ibm.com/articles/compare-deep-learning-frameworks/ (Erişim Tarihi: 15.09.2023).
45. Huang, G., Liu, Z., van der Maaten, L., ve Weinberger, K.Q. Densely connected convolutional networks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 4700–4708.
46. Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J. ve Wojna, Z. Rethinking the inception architecture for computer vision. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2818-2826. 2016.
47. He, K., Zhang, X., Ren, S., ve Sun, J. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778. 2016.
48. Chollet, F. Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 1251-1258. 2017.
49. Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., ve Chen, L-C. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 4510-4520. 2018.
50. Powers, D. Evaluation: From Precision, Recall and F-Measure to ROC, Informedness, Markedness & Correlation. Journal of Machine Learning Technologies 2.1 (2011): 37-63.
51. Powers, D. M. W. Evaluation: from precision, recall and F-measure to ROC, informedness, markedness and correlation. arXiv preprint arXiv:2010.16061 (2020)