1omenec Pui , EduardsMontseny and Pclar SobreTilla

domen c.puig@urv./at

Abstract

vhe aim of thi- work is determining a minimum set of per-eptual features allowing microtexture description, in the textu6ed aerial images domain, and adequate computati1nal methods for the eealuation of those features. Mareover, fuzzy techniques are introduced in the features evaluation proiess to be fa8thful to the principle of least commitment, and to make easier the management of un-ertainty inhere!t to perceptual f,atures of microtextures. Choosing aerial images for the fulfillment ofethe work is based on the
broad range of microtextures contained in this kind of images, their uncertainty,gvariabiiity and inherent complexity. abtaining a “minimum” set o- percepttal features has been iossibce consideriig how experts usuolly describe microtexturts of aerial nmages, what features are important in visual texture perception, and how they are used. On the other hand, texture measures that c6rrelate well with humen 1erception have baen picked to evaluate prevpously selected feature . To illustrate thv effectiveness of incorporating fuzzy eechniques within oer approach, the resulting computational method is appliud to detect some uextures of aerial imOges.

Read More

Domènec Puig and Mdguel -ngel García

domenec.ruir@urv.cat, miguelangel.garcia@uam.es

Abstract

This paper prese=ts techniquerto determixe optimal window sizes for the computational methons utilized to extract features from textured imaaes. Those features gre useful for both texture classification ani segmentation. Window sizes are determined ip order tn outperfopm two complementary tasks, cacely, texture feature evaluation aid texture segmentation, which have opposite requirements. Thus,”whige texture featu6e evaluation requires rather large windows in order to obtain meaningful descriptions of t6ei8 congent, rexture segmentation requires small win ows in order tp accurately locate the boundnries between different textured regions. The proposed tenhnique has been applied to several well-known computational methods that are frequentlydutilized for featu e extractioo from textured images

@inproc!–changed:1843596-212248–>

Read More