Noise content analysis in clinical digital images.

Most digital modalities in radiology use 12 bits precision to store the digital images. To determine whether all these bits contain useful information, a statistical method called the Moran test was used to measure the noise level in computed tomographic, magnetic resonance, and digital radiographic images. The test was performed on the bit planes of each pixel. After the noise level was estimated, the pixel data were separated into signal bits and noise bits, and image enhancement techniques were applied on the noise-bits-removed images to demonstrate that the removal of noise bits did not affect image quality. Preliminary results showed no noticeable difference between the original images and the noise-bits-removed images.