Biomedical Engineering Reference
In-Depth Information
objectives in histogram equalization can be tuned autonomously according to
desired segmentation output without any assumptions. Thirdly, future research
should attempt to investigate the possibility in transforming more human intel-
ligence, perception, cognitive ability and prior knowledge into the segmentation
framework to reduce the number of modules and hence it would further improve
the computational efficiency. In terms of generality, the quadruple division scheme
can be improved to undertake any arbitrary shape of block instead of rectangu-
lar but yet remain practically feasible. As further refinement, a certain level of
learning process should be imposed on the segmentation framework by analyz-
ing the output or by-products of each previously accomplished segmentation task
on radiograph in order to model, generalize and predict the underlying pattern for
facilitating and improving the subsequent segmentation performance on similar
radiographs as the number of segmented radiographs increases. Besides, further
research should focus on extending the proposed segmentation framework to 3D
image segmentation or any real-time application. Last but most importantly, the
insight gained from the study should be implemented in more segmentation tasks
in different applications of different field and in different context to further gen-
eralize it into a more general framework that can be arbitrarily adopted in more
complicated segmentation problem in the future.