Processing of low-intensity CCD images for temperature mapping in industrial processes

The temperature-mapping system presented is based on a grayscale CCD camera, rotating near infrared filters and an image processing software which is used for thermal imaging of heated metal parts in heat treatment processes (plasma nitriding). Due to relatively low process temperatures of about 300°C - 500°C the received light intensity of the glowing metal parts is very weak. The quality of these images can be improved by applying modern image processing filter methods. A precondition for a correct application of these denoising filters is an exact noise characterisation of the imaging system. A camera characterisation procedure based on the photon transfer technique and the system transfer function is presented. Based on these results, examples for density and intensity estimations of Poisson noise images with multiresolution methods (Platelets) are presented. The results are used to improve image quality and measurement accuracy of CCD based thermal images.