This paper describes the Bayesian image reconstruction algorithm with entropy prior with space variant hyperparameter. The spatial variation of the hyperparameter allows different degrees of resolution in areas of high and low signal/noise ratio, thus avoiding the large residuals present in algorithms that use a constant balancing parameter. The space variant hyperparameter determines the relative weight between the prior information and the likelihood defining the degree of smoothness of the solution. To compute the variable hyperparameter we used a segmentation technique based on artificial neural networks of the Self- Organizing Map type. Using this technique we segmented the image in 25 regions and computed a different value of the hyperparameter for each one. We applied the method to the Hubble Space Telescope Cameras and to ground based CCD data.