REGION SPLITTING OF MEDICAL IMAGES BASED UPON BIMODALITY ANALYSIS
Dane P. Icoffke · 1990
In this study we investigate a method for segmenting medical images into homogeneous regions via a splitting process guided by bimodality analysis. A regioii is split into four non-overlapping regions if it is found to be bimodal. Otherwise all the pixels’ gray scale values (gsv) in the region are set to the mean of the region. The decision to subdivide is made based upon a comparisoii of a bimodality measure of the region to a threshold. By varying the value of the thresholds throughout the splitting process, different degrees of segmentation coarseness can be obtained. The approach was applied to a ventricular cardiogram using six different sets of thresholds to show the effect that various threshold values had on the final result.