Soil Science Tomographic Projections Filtering using Discret Kalman and Neural Networks

In this work the use of Kalman filtering is considered to treat tomographic projections, gotten for the mini-tomographic for Soil science developed by EMBRAPA Instrumentaco Agropecuria, disturbed for a time and space variant noise, to get an improvement in the relation signal-noise. The results validation was made using ISNR (improvement in signal- noise ratio) as well as the details loss of produced images, which had been generated with the use of the filtered retroprojection algorithm. Is presented a Kalman filtering modification to treat the degraded projections with Poisson noise. To a best ISNR, an artificial neural network had been use in set with the filter.