A comparison of visual assessment and digital fractal texture analysis of bread-crumb features

Abstract

Bread is inherently irregular in conformation and it is not straightforward to quantify its physical structure. With digital image analysis, meaningful features of textured surfaces can be obtained using spectral and fractal methods. The objective of this work was to compare some bread crumb features obtained using fractal analysis with visual characteristics obtained by panellists. Ten trained panellists were asked to rate 140 bread-crumb images (520 X 520 pixels) on number of cells, uniformity, and orientation, using continuous line scales. Several methods to estimate fractal dimension of surface intensity were programmed in Matlab 6.1. The images that were rated as moderately to highly orientated by panellists presented a high textural orientation in their power spectra in the range of 90-135 degrees. The number of cells (visual) was better correlated to the fractal dimension by the method of relative differential box-counting (R=0.863) followed by the method of fractional Brownian motion (R=0.750). The range of coarsest to finest crumb corresponded to a fractal dimension of 2.36 to 2.54. The degree of crumb uniformity (visual) showed good correlation to the mass fractal dimension of the pore space (R=0.853) followed by the frequency domain method (R=0.747). In summary, the fractal texture analysis is a suitable method to assess the grain quality of bread and to characterise such uneven structures including pores, heterogeneity, and apparently replicating structures.