Publications

Abstract : In this study, a replica based image reconstruction is designed to provide high-quality reconstructions from very sparse space data. The problem of reconstructing an image from its unequal frequency samples take place in many applications. Images are observed on a spherical manifold, where one seeks to get an improved unidentified image from linear capacity, which is noisy, imperfect through a convolution procedure. Existing framework for Total Variation (TV) inpainting on the sphere includes fast methods to render the inpainting problem computationally practicable at high-resolution. In recent times, a new sampling theorem on the sphere developed, reduces the necessary number of samples by a feature of two for equiangular sampling schemes but the image that is not extremely sparse in its gradient. Total Variation (TV) inpainting fails to recover signals in the spatial domain directly with improved dimensionality signal. The regularization behavior is explained by using the theory of Lagrangian multiplier but space limitation fails to discover effective connection. To overcome these issues, Replica based Image Reconstruction (RIR) is developed in this study. RIR presents reconstruction results using both Cartesian and spiral sampling strategies using data simulated from a real acquisition to improved dimensionality signal in the spatial domain directly. RIR combined with the Global Reconstruction Constraint to remove the noisy imperfect area and highly sparse in its gradient. The proposed RIR method leads to significant improvements in SNR with very sparse space for effective analytical connection result. Moreover, the gain in SNR is traded for fewer space samples. Experimental evaluation is performed on the fMRI Data Set for Visual Image Reconstruction. RIR method performance is compared against the exiting TV framework in terms of execution time, Noisy area in SNR, accuracy rate, computational complexity, mean relative error and image dimensionality enhancement.