A Run-length Based Algorithm for Feature Extraction from Multi-target Image

This paper presented a new run-length encoding based method of extracting features from multi-target image. The algorithm uses linked list structure to store data. The data are comprised of three parts, the message in run length, the equal values and the features of the image, which are in one-to-one mapping. While scanning, we only record the features of the image according to the label. After scanning the last pixel, we merge the linked list which contains the features of image according to the equal list. Compared with other approaches, the proposed one is faster and simpler, especially it can label the binary image exactly, no matter how complicated the image is.