This paper proposes a new single-scan algorithm for segmenting the independent regions from binary images, evaluating their boundaries, and coordinates of centroids. Its main benefit, compared to the known solutions, is that it does not require changing the original image or creating a new one with a similar size to store labeling information. Particular attention in the development of the algorithm is paid to reducing both the amount of memory needed to implement it and the number of records in memory during the scan. This makes it perfect for further FPGA implementation.