This thesis is concerned with the problem of colour shade grading for Industrial Inspection and attempts to find accurate and robust solutions to this problem. The application we are interested in, is the automation of the ceramic tiles manufacturing process so as to replace the human inspectors responsible for the quality control of the product. Therefore our aim is to perform the colour grading in a way which is consistent with what the human experts and subsequently the clients would perceive. First an overview of colour vision, colour measurement and colour constancy is given. Then a method that tackles the problem of colour grading of uniform and patterned surfaces is proposed. This method is the first step towards colour grading since it involves various corrections of the data, so as to provide the necessary precision for any further attempt. The problem of colour grading of random textures is then addressed. A method based on the comparison between colour histograms is proposed, and various statistical aspects involved in the comparison of distributions such as the colour histograms are discussed. Since the real-time implementation of any industrial inspection method should be taken into account, we use a space-effective method of storing colour histograms. Having solved the problem of colour grading for the majority of uniform and textured surfaces, we then try to optimise the performance of the proposed techniques, for cases where it fails. We attribute that to the fact that every electronic sensor captures colour and patterns in a way which only approximates what the human vision system would perceive. First we propose a method of perceptual colour grading of uniform surfaces, which transforms the camera data to data as they would have been recorded by the human eye. This method makes use of metameric data, to determine the relation between the human and the electronic sensors. We use various methods of generating metamers, and we show how the need of a spectrophotometer can be overcome. In a similar way, we propose a method of perceptual colour grading of random textures, which involves the restoration of the electronically acquired data and then their transformation to a colour space which expresses the way we perceive colour texture. We test both methods with real data, and we compare them with the non-perceptual ones. All the methods proposed in this thesis have been tested with real data, from the ceramic tiles manufacturing industry, previously colour graded by human inspectors. The consistency of the methods has been tested by using various sets of all sorts of tiles, and by repeating the acquisition and grading processes many times for every set of tiles. Further, these experiments have been carried out using different apparatuses, thus allowing us to draw conclusions about their quality and to make our methods as hardware independent as possible.