Abstract #3185

Plett J, Lillo I, Guarini M, Irarrazaval P

MRI images can be reconstructed reasonably well from just a few randomly chosen k-space samples through a convex L1 optimization. We present a new alternative for the domain of the optimization function based on the Discrete Cosine Transform (DCT) and compare it to Wavelets. The results indicate better and faster performance for the DCT algorithm reducing reconstruction errors and increasing Peak Signal to Noise Ratio (PSNR).