Microstructure segmentation using active contours-Possibilities and limitations

Image data often carries redundant or unnecessary information. Consider a hypothetical image containing a phase embedded into a matrix. If the shape of the phase is of interest then the matrix information is redundant. Identification and classification of the pixels belonging to the matrix and the phase is a process called image segmentation. Active contouring is an edge detection based segmentation method that produces results superior to that of traditional edge detectors. High noise tolerance and built-in functional flexibility offered by the active contouring models make it an attractive technique in various applications including data compression, tracking of growth of a phase, and boundary outlining for further three-dimensional volume reconstruction. In this paper, the factors impacting the final contours will be discussed and shown in practice. It will also be shown that the technique is especially attractive when employed to process complex real microstructure datasets with limited user interaction.