Image decorrelation: a procedure for the study of X-ray images in medical studies

The authors propose a new concept known as neighbourhood images to generate a sequence of images from dilation to erosion and from closing to that of opening on the basis of an algorithm presented by K. Vladimir (1986) in the morphologic domain. These generated images show a high correlation to each other. Basically, the morphologic procedures separate the border regions from those of the interior and as such, the image content and border component are affected in the morphologic transformation. The authors further utilised the dilated, the median and eroded images for decorrelation and generation of images which are ortho-normal to each other. Such decorrelated images can be utilised for the study of the borders separately from that of interior pixels. Based on this background, X-ray image of the neck and chest region displaying the malignancy near the neck region is studied. Neighbourhood images are selected and a set of three morphologic images are used for decorrelation. The first image shows enhancement with reduction of noise content. The second and the third image display the contours of the malignancy. X-ray images are routinely used, for the visual interpretation and analysis, and are not subject to rigorous procedures of image analysis