Approximate statistical properties of reconstructed images using model-based bootstrapping

In tomographic image reconstructions, it is often necessary to compute certain statistics of a reconstructed image. These quantities can be used, for example, to analyze the noise properties of a reconstruction. This paper introduces the model-based bootstrapping method to approximate mean and variance-covariance matrix of a reconstruction. This approximation is versatile and can be implemented in different tomographic reconstructions, such as emission and transmission tomographies. This paper also considers the possibility of computational load reduction by a simultaneous multiplicative iterative reconstruction algorithm.