A new methodology to solve the problem of characterizing 2-D biomedical shapes.

In this paper, a new approach to solve the problem of characterizing 2-D biomedical shapes is introduced. Two-dimensional biomedical contours are described through a 'degrees of smoothing' vector in which each component determines the proper degree of detail for representing each curve part isolating a single structure. A segmentation process is designed based on a clustering procedure applied to vectors of texture measures which are obtained from the graph of curvatures. To solve the problem of characterizing biomedical shapes, a suitable interpolation procedure between the most outstanding perceptual points from the smoothed contours is given.

Line detection and texture analysis for automatic nematode identification.

This paper is the second in a series studying procedures for estimating and calibrating features of nematodes from digital images. Two kinds of features were analyzed for recognition: those with a directional component and those with a textural component. Features that have a directional component (lateral field and annules) were preprocessed with classic algorithms and modified by directional filters. Features having texture (esophagus and intestine) were analyzed with vectors of measures to define them and the statistical technique CART (classification and regression trees) to explain the role that each measure plays in the identification and discrimination process.