Control of ham salting by using image segmentation A.J. Sa ´nchez a, * , W. Albarracin b , R. Grau b , C. Ricolfe a , J.M. Barat b a Departamento de Ingenierı ´a de Sistemas y Automa ´ tica, Universidad Polite ´cnica de Valencia, Spain b Departamento de Tecnologı ´a de Alimentos, Universidad Polite ´cnica de Valencia, Spain Received 16 November 2006; received in revised form 7 February 2007; accepted 16 February 2007 Abstract Curing is one of the most traditional processes in the meat industry, being used in a great variety of products such as cured ham. During the salting process the raw material acquires the curing agents, allowing the safe development of subsequent stages in processing. Digital image analysis has been used in different food research areas. Most of the studies that use image analysis in the evaluation of different aspects of meat products have been carried out mainly on ham, detecting quality problems of the product. However, none of these studies deals with the influence of different components present on the ham surface (fat, connective tissue and lean) and its rela- tionship with mass transfer during ham processing. The aim of this study was the use of image segmentation to quantify the lean, fatty and connective tissue areas on the ham surface and determine the relationship of those areas to salt gain during the salting process. The obtained results show that image segmentation algorithm can be used in combination with other parameters values to predicting ham behaviour during the salting process. Ó 2007 Elsevier Ltd. All rights reserved. Keywords: Image segmentation; Ham salting; Mass transfer areas 1. Introduction Machine vision is a novel technology for acquiring and analyzing images of a real scene in order to extract any required information (Davies, 1997). Several steps are involved in the image analysis process (Gonzalez & Wints, 1991). The first step in the analysis process is the acquisition of the image, in which colour images are usually captured and represented by the primary colours red, green and blue (RGB space). Different colour spaces can be obtained from the RGB space using known equations. The CIE (Commis- sion Internationale de L’Eclairage) derived and standard- ized two perceptually uniform colour spaces from CIE XYZ: the CIE Luv and the CIE Lab. These spaces could be of interest for analyzing the main factors of human col- our perception (Hunt, 1991). The second step is image pre-processing, where the qual- ity of the image is improved. Uncertainty in the captured data is a generic problem in analysing scenes containing food objects in semi-controlled light conditions. Therefore some pre-processing techniques can be applied in order to reduce the uncertainty of the captured data. The third step is image segmentation, in which the image is clustered into regions of interest (foreground and back- ground regions). In general, automated segmentation is one of the most difficult steps in image analysis. Food image segmentation is still an unsolved problem because of its complex attributes and constraints. Finally, the ability to perform pattern recognition at some level is fundamental to image analysis. A pattern is a quantitative or structural description of an object of interest in an image. Pattern recognition by machine involves techniques for assigning patterns automatically to their respective classes. In machine recognition of image 0956-7135/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodcont.2007.02.012 * Corresponding author. Tel.: +34 9638 77000x75786; fax: +34 9638 79579. E-mail address: asanchez@isa.upv.es (A.J. Sa ´nchez). www.elsevier.com/locate/foodcont Food Control 19 (2008) 135–142