A novel image registration hyper spectral classification method for multi-scale image segmentation based on morphology and attributes

In this paper a novel image registration hyper spectral classification method based shape features and multi scale image segmentation is introduced. In this initially two images are taken one is image 1 and another one is reference image. Now to this input images multi scale segmentation is applied based on convex model. After applying segmentation the image is segmented into various parts. Now from the segmented part over ranged areas are removed. Image registration will be done based on the global spatial relation technique. Hyper spectral image is obtained and features are extracted. After features are extracted classification process is done, initially samples are trained and then classification is done by using PCA wavelet transform technique. By using C50 algorithm decision tree classification is applied. At last classified image is obtained with high accuracy and high quality of assessment. From experimental results it can observe that the accuracy, quality, precision, F1 score is increased and error rate is reduced.