Recursive Implementation of Erosions and Dilations Along Discrete Lines at Arbitrary Angles
IEEE Transactions on Pattern Analysis and Machine Intelligence
The watershed transform: definitions, algorithms and parallelization strategies
Fundamenta Informaticae - Special issue on mathematical morphology
Level Set Segmentation Using Statistical Shape Priors
CVPRW '06 Proceedings of the 2006 Conference on Computer Vision and Pattern Recognition Workshop
IEEE Transactions on Signal Processing
Image enhancement and denoising by complex diffusion processes
IEEE Transactions on Pattern Analysis and Machine Intelligence
Hi-index | 0.00 |
Automatic, high-throughput, quantification of the precise position and orientation of biological objects is essential for studying living, biomedically relevant processes from time-lapse microscopy images. These measurements frequently include precise estimates for the center-of-mass as well as the location of the true object boundaries e.g. the membranes of cells. This paper describes a region-oriented segmentation approach applied to the detection of both insects at the mm length scale as well as bacteria at the µm length scale. Despite the differences in length scale, images of both objects have similar aspect ratio, and it is common to have overlapping objects in images of both. This thus presents a challenge for any segmentation algorithm. Our approach performs all orientation detection through a chord length transform, so the task of separating overlapping objects in a two-dimensional image is reformulated as a voxel-labeling problem within a three-dimensional volume. It then utilizes the directional information from the Radon transformed image. Experimental results in simulation show that our method is effective in separating clustered elongated but stubby objects with aspect ratios not far from 1. The applications in detecting insects and Escherichia coli bacteria demonstrate the value of our approach.