We present in this work a new method for the tomographic reconstruction with Poisson noise corrupted projections. The reconstruction method is a hybrid estimation theoretic-POCS (Projection Onto Convex Sets) method performed by first estimating the noisy projections using Anscombe Transformation, Discrete Wavelet Transform and pointwise Wiener filter and subsequently reconstructing the images through a parallel POCS method. The use of a previous filtering of the projections imposes a tighter restriction on the observation of the POCS method, allowing a better reconstruction, with a very little increase in computational cost.