Approaches to improve accuracy of neural network classification of images dominated by mixed pixels.
Mohamed A. IbrahimManoj K. AroraSusmita K. GhoshHua-mei ChenPublished in: IGARSS (2004)
Keyphrases
- input image
- neural network
- classification accuracy
- image classification
- discriminative classifiers
- image pixels
- pixel values
- intensity values
- image data
- image features
- pattern recognition
- original images
- image database
- pixel wise
- image regions
- ground truth
- classification rate
- image registration
- neural network is trained
- learning vector quantization
- image analysis
- reference images
- support vector machine
- segmentation method
- gray level images
- linearly combined
- multi layer perceptron
- pixel classification
- image set
- test images
- spatial information
- neighboring pixels
- landsat etm
- grey level
- hyperspectral images
- support vector machine svm
- edge detection
- feature vectors
- artificial neural networks
- image retrieval
- feature space
- decision trees
- spatial correlation
- feature maps
- pixel intensities
- gray value
- aspect ratio
- training set