Gaussian bilateral filtered discrete Hartley feature transformation based infomax boosting for hyperspectral image classification
Hyperspectral Image (HSI) is a spatially sampled image collected from a lot of neighboring narrowed spectrum bands via hyperspectral sensor. The process of HSIs is a difficult obligation because of the high-dimensional image features. Conventional techniques mainly utilize hand-made characteristics...
Guardado en:
| Autores principales: | , |
|---|---|
| Formato: | article |
| Lenguaje: | EN |
| Publicado: |
KeAi Communications Co., Ltd.
2021
|
| Materias: | |
| Acceso en línea: | https://doaj.org/article/f2ecc5e5b8af4a459ae2bde5a638e495 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | Hyperspectral Image (HSI) is a spatially sampled image collected from a lot of neighboring narrowed spectrum bands via hyperspectral sensor. The process of HSIs is a difficult obligation because of the high-dimensional image features. Conventional techniques mainly utilize hand-made characteristics that are not fit for classification. For the purpose of enhancing the correctness and reducing the overall time complexity, an efficient technique called Gaussian Bilateral Filtered Hartley Feature Transformation Based Infomax Boost Classification (GBFHFT-IBC) is introduced. The input hyperspectral image is divided into the number of spectral bands. The proposed GBFHFT-IBC method includes three fundamental procedures, which include preprocessing, characteristics classification and extraction. Preprocessing is carried out using a Gaussian kernelized bilateral filter that replaces the pixel intensity from the picture with the weighted to take the position of the intensity of every pixel from the picture with the prejudiced average of neighboring pixels resulting in it removes the noisy pixels and improve the image quality. After preprocessing, the different characteristics like texture, color and shape are extracted from the image using the Discrete Hartley transformation technique. With the extracted features, the Infomax Boost method is applicable when classifying different scenes from the spectrum band, which has to be grouped into various categories with the top categorization accuracy that consider the minimal error rates. The experiments are done using the hyperspectral picture dataset composed of various factors like categorization accuracy, PSNR, positive and negative rate, and timeline complexity considering the number of spectral bands. The observed result shows that the presented GBFHFT-IBC method improves the hyperspectral image classification correctness, PSNR and minimized negative and positive rates, which include timeline complexity compared to using the sophisticated techniques. |
|---|