A segmentation and object extraction algorithm with linear memory and time constraints

An experimental segmentation and object extraction algorithm is described. The system was developed for medical image processing with the primary application being DNA (deoxyribonucleic acid) sequencing. A typical DNA sequencing can involve processing the image of an autodiagram of size 14×17 inches resulting in a 2048×8600 digitized image under the specified spatial resolutions. The digitized image is too big to manage, even using super-minicomputers such as DEC VAX 11/780, and to perform any amount of classical image processing. Therefore, an elegant hardware and software design is necessary to deal with the large image and to complete the image-understanding task in an efficient manner. This work focuses on the image-processing aspects of the system and describes the run-length image representation, a link list data structure, a heuristic connected component analysis algorithm based on the data structure, a primitive object segmentation algorithm, and feature extraction