<title>Detection of skin line in computed radiographs for improved tone scale</title>

Automatic Detection of Radiation Fields in Digital Radiographic Images

“…If direct exposure is detected, a two-pass approach is used. In the first pass the method uses the classifier of the skin line detection process 13 to identify direct-exposure-collimation boundary transitions. This classification process involves the computation of the set of features described in Table 1.…”

“…Computation of the features is followed by application of a Gaussian maximum likelihood classifier. 13 Since the boundary transitions found in this first pass are the most reliable, the pixels associated with the maximum gradient of these transitions are labeled with a count of 255 ("high likelihood"), which corresponds to a collimation-direct exposure, or foregroundbackground (later referred to as F/B) transition. The results of this first pass (in both orientations) are also used to establish an estimate of the maximum gray level of the collimation using the histogram of the beginning points of the direct-exposurecollimation boundary transitions.…”

“…The main motivation for such a three-stage scheme is from the past experiences in related work. 10,13 It has been found that the edge profile of the collimation boundary and the skin line show quite distinctive characteristics. Edge profile analysis can reliably identify whether an edge pixel sits on the transition from the shadow region to the irradiation field, or on the skin line (i.e.…”