Description of the kinetic enzymatic browning in banana (Musa cavendish) slices using non-uniform color information from digital images Roberto Quevedo a, * , Oscar Díaz a , Betty Ronceros a , Franco Pedreschi b , José Miguel Aguilera b a FITOGEN Program, Department of Science and Food Technology, Universidad de Los Lagos, Av. Fushlocher 1305, Osorno, Chile b Department of Chemical Engineering and Bioprocesses, Pontificia Universidad Catolica de Chile, P.O. Box 306, Santiago, Chile article info Article history: Received 29 October 2008 Accepted 7 April 2009 Keywords: Enzymatic browning Banana slices Computer vision systems Fourier fractal texture abstract A novel methodology ‘‘fractal browning indicator” (FBI) is presented, that describes the enzymatic browning kinetic based on the use of irregular color patterns from banana slice images. It uses the fractal Fourier texture image value in a selected area, to calculate a fractal dimension (FD), which represents the complexity of color distribution. During the procedure, colors from digital images were first transformed to L  a  b  space color using a transformation function (quadratic model), in order to derivate three color channels, lightness (L  ), redness (a  ), and yellowness (b  ). In the results, lightness and yellowness param- eters decreased during the browning kinetic, when their respective FD values increased, indicating major color distribution complexity in a selected area analyzed during the kinetic. The redness color (a  ) did not show any statistical variation. The empirical power law model was suitable to correlate enzymatic browning kinetic data both for FBI and for the traditional method (when an L  mean was used). However, enzymatic browning rates using the FBI method, were between 8.5 and 35 times higher than rates cal- culated with the traditional method. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction Browning of raw fruits is a major problem in the food indus- try and is believed to be one of the main causes of quality loss during post-harvest handling and processing. In fact, when fruits are cut, the cut surface turns brown; it reduces not only the vi- sual quality but also results in undesirable changes in flavour and nutrient loss due to enzymatic browning (Luo & Barbosa, 1997). Browning can cause deleterious changes in the appear- ance and organoleptic properties of the food market, value, and in some cases, complete exclusion of the food product from certain markets (McEvily, Iyengar, & Otwell, 1992). The control of cut-surface browning is critical to maintaining the quality and safety of fresh-cut produce. Traditionally, enzymatic browning has been quantified using browning indicators trough of biochemical index, for example the polyphenol oxidase activity (Hosoda et al., 2005; Murata, 2001; Osanai, Motomura, & Sakurai, 2003; Sannomaru, Katayama, Kashimura, & Kaneko, 1998; Sharon & Kahn, 1979; Waliszewski, Pardio, & Ovando, 2007), or physical indicators, such as the color surface (Elshimi, 1993; Kang et al., 2004; Lambrecht, 1995; Lozano de Gonzalez, Barrett, Wrolstad, & Durst, 1993; Lozano, Drudisbiscarri, & Ibarzribas, 1994; Lu, Luo, Turner, & Feng, 2007; Luo & Barbosa, 1997; Molme, Buta, & Newman, 1999; Shengmin, Yaguang, Ellen Turner, & Hao Feng, 2007). In the case of physical indicators based on color, L  a  b  space or CIELab has been the most extensively used color model; especially the L  value, which has been used as a browning indicator in fruits (Luo & Barbosa, 1997, 1994, 1995; Luo & Patterson, 1994; Parpinello, Chinnici, Versari, & Riponi, 2002; Pristijono, Wills, & Golding, 2006; Sapers & Douglas, 1987; Sapers & Ziolkowski, 1987; Severini, Baiano, De Pilli, Romaniello, & Derossi, 2003; Soliva-Fortuny, Elez- Martinez, Sebastian-Caldero, & Martin-Belloso, 2002; Valentines, Vilaplana, Torres, Usall, & Larrigaudiere, 2005). CIELab or L  a  b  space color is device-independent, creating consistent colors regardless of the device being used to acquire the image. L  is the luminance or lightness component, it ranges from 0 to 100, while a  (green to red, or redness) and b  (blue to yellow, or yellowness) are two chromatic components, with values varying from 120 to +120 (Papadakis, Abdul-Malek, Kamdem, & Jam, 2000). Recently, Leon, Mery, Pedreschi, and Leon (2006) have suggested a computer vision system (CVS) to measure color in L  a  b  from RGB space to be used in image analysis and some work using that approximation has been applied in food (Pedreschi, Bustos, et al., 2007; Pedreschi, Leon, et al., 2007; Quevedo, Aguilera, & Pedreschi, 2008). During the description of browning kinetic in bananas, using color information, a color mean value is generally assumed for the region analyzed. That is, an average of the L  , a  or b  values is calculated from colorimeter devices or CVS in the area analyzed 0963-9969/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodres.2009.04.004 * Corresponding author. Fax: +56 64 333226. E-mail address: rquevedo@ulagos.cl (R. Quevedo). Food Research International 42 (2009) 1309–1314 Contents lists available at ScienceDirect Food Research International journal homepage: www.elsevier.com/locate/foodres