A technique to correct outlier pixels in images acquired with CCD-based cameras

Images acquired with CCD-based cameras usually contain some spots or regions which do not match with their neighbourhood, even when focusing a homogeneously illuminated surface of homogeneous colour. This behaviour, caused by different responses of the CCD elements to red, blue and green light, is specially noticeable in the case of high sensitivity cameras where even in complete darkness some bright spots still appears. A technique has been developed in this work to correct defective or outlier pixels in the stored image, making thus possible to retrieve a cleaner and improved picture afterwards. It is based on the determination of a matrix of defects, obtained from the camera response to complete darkness, after a criterion which takes into account the deviation of red, green and blue (RGB) values from the corresponding average values occurring in a selected region of acceptable homogeneity. When the camera operates in the normal fashion, focusing an illuminated object, the defective pixels would exhibit naturally other RGB values. Yet, as a systematic error, the matrix of defects keeps its pattern, and hence, the defective pixels, even if not clearly identifiable in the image, are obviously unreliable. Therefore, their RGB values are replaced by average numbers computed from the surrounding pixels, excluding those ones which are themselves, as well, defectives. All these operations are automatically carried out by an algorithm written in Fortran language, fed by matrices containing the images to be cleaned up, and the matrix of defects