A run-based algorithm for identifying connected components in binary images

We present a run-based algorithm for identifying connected components in binary images. We define affecting scope of a run and use an array whose size is equal to width of the image to store runs' labels in a row. The label's storing scope in the array is the run's affecting scope. By checking values in the array we can know whether the current run connects with any upper-row neighbor runs. We also define updating principles to maintain the array. The algorithm only needs to scan image one pass and need not maintain an equivalent table, so its speed is fast and it is very simple. It can identify 4-connectivity and 8-connectivity connected components in binary images. Experiment results show that the algorithm is effective.