A two-dimensional extension of minimum cross entropy thresholding for the segmentation of ultrasound images

Segmentation is often an important step in medical image analysis. The local entropy is a possible variable for segmenting ultrasound images containing fluid surrounded by a soft tissue. A commonly used tool for image segmentation is thresholding. Recently, a new thresholding technique, known as "minimum cross entropy thresholding" (MCE), has been proposed. We present a multivariate extension of MCE in which the segmented variable (gray level) is replaced by a weighted combination of several image parameters. We propose to use a bivariate extension of MCE, which uses a linear combination of the gray level and the local entropy. The results obtained are demonstrated for ultrasound images of ovarian cysts.