Noise reduction in spine videofluoroscopic images using the undecimated wavelet transform.

Videofluoroscopy permits using sequences of low quality images to study the spine movement. In this work the problem of enhancing the quality of these images is considered in order to facilitate the extraction of kinematic parameters. The method is based on the undecimated wavelet transform and on a preliminary training of a sub-set of images. The anatomical features are preserved using a mask. Key element of the method is its fast and automated implementation. The concept of improving the extraction of kinematic parameters by improving the image representation instead of the technique to extract these is also innovative. The technique has been tested on two sequences of images and the results demonstrates that the method enhances images not related with training sub-set.