Improved Registration of Infrared Images Using EOH Descriptor
| Autor: | P. Sita Sowjanya, J. Prasanna Kumar, B. Sandhya |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Scale (ratio)
Computer science business.industry Geometric transformation ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Scale-invariant feature transform Image registration Feature (computer vision) Computer Science::Computer Vision and Pattern Recognition Histogram Computer vision Noise (video) Artificial intelligence business Rotation (mathematics) |
| Zdroj: | Advances in Computational Intelligence and Informatics ISBN: 9789811533372 |
| DOI: | 10.1007/978-981-15-3338-9_13 |
| Popis: | Feature-based image registration involves overlaying two images of the same area by extracting features, matching and computing geometric transformation. Multimodal image registration is useful in a variety of applications as the unique information contained in diverse images can be combined. Descriptors proposed for multimodal image matching such as edge-oriented histogram (EOH), log-Gabor histogram (LGHD) can address the photometric variation between the visual and infrared images better than conventional image descriptors such as SIFT. However, the invariance of such descriptors to geometric variations such as scale and rotation is poor. To address the geometric variations in addition to photometric variations, the region around the feature point is preprocessed using scale and rotation information of detector before deriving the descriptor. Different datasets are composed of images obtained in visible light and infrared spectra images, and IR images contain variations to compare scale, rotation, noise, blur, etc., to the performance with those of state-of-the-art algorithms. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|