Evaluation of colour spaces in computer vision application of wood defects detection

The automation of quality control for the appearance of wood by digital image processing poses the problem of defect specification. In the case of defects showing little contrast or colour, black and white vision systems are inadequate and colour vision is needed. Colour entails a three-dimensional working space, thus permitting a series or geometric transforms of greater or lesser relevance depending on the specific application in question. In this article, we develop a method which helps to select the best colorimetric transform among those currently being used. This method is then extended to the problem of defect segmentation in wood, and culminates with a colour-analysis algorithm which is effective in terms of detection quality.<<ETX>>