Three-dimensional image recognition and visualization

Date of Completion

January 2009


Engineering, Electronics and Electrical




The research is divided in two parts, the first part is three dimensional image recognition, and the second part is three dimensional image fusion. ^ Three-dimensional information is collected using holography or integral imaging or photon counting integral imaging. In the dissertation, independent component analysis successfully extracts features from the training data, separate sources as well as transform three dimensional image data into transformed domain. Problem of large dimension is tackled using principle component analysis. Classification task is based on various criteria, such as Euclidean distance, k-nearest neighbor, and cosine value.^ Discrete wavelet transform is another method for decomposition of 3D image data into transformed domain, which is used in the image fusion part. Denoising techniques have been mentioned to reduce the speckle noise in reconstructed holographic images. ^