A Method for Filling Holes in Objects of Medical Images Using Region Labeling and Run Length Encoding Schemes

This paper presents a new method for filling holes in objects of binary images. Further it analyses the performance of other existing holes filling operations in binary form of medical images. Hole filling operations are widely used in medical image processing today. Almost all the medical image processing operations produce a binary form of original image at any stage. The binary images are normally produced by simple segmentation techniques such as thresholding. They contain foreground objects surrounded by background regions. Some times a set of background regions lie completely within the foreground regions due to imperfection in the binary conversion identified by the optimal thersholding. It is known as hole within the foreground objects. Hence a hole is an area of dark pixels surrounded by light pixels. Variety of holes filling operations are developed and used by researchers for medical image analysis study. They are Areafill operation, Morphological operation, and Floodfill operation. Sometimes a combination of these methods along with image subtraction is used to identify the holes effectively. The existing operations are required user intervention either to start the process of refine the results. The user intervention is application oriented and subject specific knowledge is required to run the process. We have developed an automatic algorithm for filling holes that appear within the binary mask. The proposed approach is based on Run Length Encoded Data (RLED) of binary image and region labeling procedure. Usually in medical imaging, the scanned organs are surrounded by air in the form of dark background. Using this expert knowledge the holes within objects are identified and removed using runs of holes and image limits. The proposed operation is fully automatic and no user intervention is required at any stage. This automatic process is best suitable holes filling module in any brain image analysis pipeline.