Restoration of chest X-ray image using a priori information

A novel chest X-ray image restoration technique is proposed for improving SNR, which is an important parameter for system performance evaluation. The filter was designed using a priori information on the imaging system and the anatomical structure, and the statistics of the projected object. A priori information of the system comprised X-ray characteristics, sensor configuration, and characteristics of each sensor. The present technique simultaneously processed two images acquired in the same region by two sensors of different characteristics. This consisted of two filters in cascade. One was a dual-channel Wiener filter (DCWF) with filter parameters designed by cross-correlation between two images. The other was a dual-projection filter (DPF), in which the projection operator was designed using a priori information of chest images based on the concept of a reliable frequency region. Two images were then processed iteratively and adaptively. The DCWF improved SNR by 1-2 dB in a simulation study compared with conventional single-channel filters. This was further enhanced by 0.1-0.2 dB when the output of the DCWF was fed for an initial estimate of the DPF by adapting the filter parameters to local regions (lung, mediastinum, subdiaphragm) and structures (bone, blood vessel, bronchi) of the chest