Multi-level thresholding using image transformation

An approach to transforming an image in order to facilitate threshold selection is presented. A desirable histogram which would ease threshold selection is specified and the original image is transformed into one whose histogram has that desirable distribution. A threshold is then easily selected for the image. Experiments have been performed by specifying bimodel histograms for single-level thresholding, and by specifying multi-model histograms for multi-level thresholding. The results of these are presented.