Simultaneous Parameter Estimation and Segmentation of Gibbs Random Fields Using Simulated Annealing
IEEE Transactions on Pattern Analysis and Machine Intelligence
Unsupervised segmentation of noisy and textured images using Markov random fields
CVGIP: Graphical Models and Image Processing
The minimum description length principle in coding and modeling
IEEE Transactions on Information Theory
Signal and noise modeling in confocal laser scanning fluorescence microscopy
MICCAI'12 Proceedings of the 15th international conference on Medical Image Computing and Computer-Assisted Intervention - Volume Part I
Hi-index | 0.01 |
Confocal laser scanning microscopy (CLSM) has been widely used in the life sciences for the characterization of cell processes because it allows the recording of the distribution of fluorescence-tagged macromolecules on a section of the living cell. It is in fact the cornerstone of many molecular transport and interaction quantification techniques where the identification of regions of interest through image segmentation is usually a required step. In many situations, because of the complexity of the recorded cellular structures or because of the amounts of data involved, image segmentation either is too difficult or inefficient to be done by hand and automated segmentation procedures have to be considered. Given the nature of CLSM images, statistical segmentation methodologies appear as natural candidates. In this work we propose a model to be used for statistical unsupervised CLSM image segmentation. The model is derived from the CLSM image formation mechanics and its performance is compared to the existing alternatives. Results show that it provides a much better description of the data on classes characterized by their mean intensity, making it suitable not only for segmentation methodologies with known number of classes but also for use with schemes aiming at the estimation of the number of classes through the application of cluster selection criteria.