Options
Digital Image Conspicuous Features Classification Using TLCNN Model with SVM Classifier
Journal
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
ISSN
03029743
Date Issued
2022-01-01
Author(s)
Rastogi, Swati
P. Duttagupta, Siddhartha
Guha, Anirban
Abstract
We present a transfer learning convolutional neural network (TLCNN) model in this study that permits classification of noticeable noise characteristics from degraded images using dispositional criteria. Various digitally degraded images comprising additive, impulsive, as well as multiplicative noise are used to test the suggested approach for noise type detection techniques. We further show that our representations sum up effectively when applied to additional datasets, achieving best-in-class results. In order to use the TL approach, we selected three CNN models: VGG19, Inception V3, and ResNet50, which are profound for visual recognition in computer vision to detect correct noise distribution. In this perplexing setting, the system’s capacity to deal with the degraded image has outperformed human vision in noise type recognition. By getting noise classification performance of 99.54, 95.91, and 99.36%, while observing the nine classes of noises, the author’s testing confirmed the constant quality of the recommended noise type’s categorization.
Volume
13256 LNCS
Subjects