An Analysis of the Neural Filter Trained to Improve Quality of Images with Quantum Noise and Realization of Approximate Filter

摘要

In this paper, we propose a new method for analyzing the neural filter (NF) which is trained to improve the image quality of medical X-ray image sequences with quantum noise. The pro- posed method can analyze an unknown nonlinear system such as the trained NF by using its outputs when the input signals for analysis are fed to it. Experimental results demonstrate that the proposed method can make the characteristics of the nonlinear system clear. AIore- over, we realize the approximate filter of it using the results of its analysis. We show that the approximate filter is a good approximation of it.