Image Reconstruction from Non- Uniformly Sampled Spectral Data Alfredo Nava-Tudela AMSC 663, Fall 2008 Midterm Progress Report Advisor: John J. Benedetto Alfredo Nava-Tudela AMSC 663, Fall 2008 Midterm Progress Report Advisor: John J. Benedetto
Outline Background/Problem Statement Algorithm Database and Validation Test Results Future Work Background/Problem Statement Algorithm Database and Validation Test Results Future Work
Background Sometimes there is a need to reconstruct from spectral data an object in the spatial/time domain For example, and image from an MRI machine Sometimes there is a need to reconstruct from spectral data an object in the spatial/time domain For example, and image from an MRI machine
Problem Statement Given a two dimensional spectral data set, reconstruct an image in the spatial domain that matches as closely as possible that data set in the spectral domain
Problem Statement In a real life application, the spectral data set is generated by some physical process In our case, we generate artificial spectral data from a known high resolution image We use a down-sampled version of that image to compare the goodness of our reconstruction In a real life application, the spectral data set is generated by some physical process In our case, we generate artificial spectral data from a known high resolution image We use a down-sampled version of that image to compare the goodness of our reconstruction
The Algorithm Stage one:
The Algorithm Stage two:
The Algorithm Stage three:
The Algorithm This algorithm corresponds to the direct solution of the linear system of equations presented in the project proposal This has the drawback of having to store a potentially very big matrix This algorithm corresponds to the direct solution of the linear system of equations presented in the project proposal This has the drawback of having to store a potentially very big matrix
Validation We select from a standard set of image processing images a subset
Validation We convert the images to grayscale, in case they are in color
Validation These are the images that we feed to our algorithm Select the desired resolution for the reconstruction: 16 by 16 and 32 by 32 Higher resolutions take longer These are the images that we feed to our algorithm Select the desired resolution for the reconstruction: 16 by 16 and 32 by 32 Higher resolutions take longer
Test Results: Baboon
Test Results: Lena
Test Results: Peppers
Future Work Allow arbitrary size input images, currently only square images processed Implement algorithm that doesn’t store matrices Write C++ code, explore parallelization Explore other ways to assess goodness of reconstruction Explore different sampling geometries Allow arbitrary size input images, currently only square images processed Implement algorithm that doesn’t store matrices Write C++ code, explore parallelization Explore other ways to assess goodness of reconstruction Explore different sampling geometries
References Adi Ben-Israel and Thomas N. E. Greville. Generalized Inverses. Springer- Verlag, John J. Benedetto and Paulo J. S. G. Ferreira. Moderm Sampling Theory: Mathematics and Applications. Birkhauser, J. W. Cooley and J. W. Tukey. An algorithm for the machine computation of complex Fourier series. Math. Comp., 19: , E. H. Moore. On reciprocal of the general algebraic matrix. Bulletin of the American Mathematical Society, 26:85-100, Diane P. O’Leary. Scientific computing with case studies. Book in preparation for publication, Roger Penrose. On best approximate solution to linear matrix equations. Proceedings of the Cambridge Philosophical Society, 52:17-19, Adi Ben-Israel and Thomas N. E. Greville. Generalized Inverses. Springer- Verlag, John J. Benedetto and Paulo J. S. G. Ferreira. Moderm Sampling Theory: Mathematics and Applications. Birkhauser, J. W. Cooley and J. W. Tukey. An algorithm for the machine computation of complex Fourier series. Math. Comp., 19: , E. H. Moore. On reciprocal of the general algebraic matrix. Bulletin of the American Mathematical Society, 26:85-100, Diane P. O’Leary. Scientific computing with case studies. Book in preparation for publication, Roger Penrose. On best approximate solution to linear matrix equations. Proceedings of the Cambridge Philosophical Society, 52:17-19, 1956.