Wiener based spatial resolution enhancement of MRI sequences of the vocal tract: A comparison between two correlation models

• Published in: 2012 19th IEEE International Conference on Image Processing pp. 869 - 872
• Main Authors: Martins, A. L. D., Mascarenhas, N. D. A.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-09-2012
• Subjects: Articulatory synthesis; Computational modeling; Correlation; Magnetic resonance imaging; Numerical models; Spatial resolution; spatial resolution enhancement; Speech; super resolution image reconstruction
• ISBN: 1467325341; 9781467325349
• ISSN: 1522-4880; 2381-8549
• DOI: 10.1109/ICIP.2012.6466998

Articulatory Synthesis consists in reproducing speech based on models of the vocal tract and of articulatory processes. Recent advances in MRI allowed for important improvements with respect to the speech comprehension. However, one of the main challenges is the high-quality acquisition of images. Since adopting more powerful acquisition devices might be financially inviable, a more feasible solution is to perform the resolution enhancement using only image processing techniques. In a previous work, we proposed a Wiener based super resolution technique for the spatial resolution enhancement. A separable Markovian model was used in the characterization of the spatial correlation structures. In this paper we report on a study that compares this model with an isotropic model. According to the conducted experiments, the isotropic model outperformed the separable Markovian model in all cases. Moreover, the adopted super resolution method out-performed the two available approaches used in this context.