The Study of 2-D Connected Component Algorithm

• Main Authors: Wei-Shin Tang, 唐偉訓
• Other Authors: Jan-Ray Liao
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/34422274243220754107

碩士 === 國立中興大學 === 通訊工程研究所 === 101 === Connected-component labeling is a fundamental algorithm in image processing. In a binary image which has been classified as foreground and background pixels, the purpose of the connected-component labeling is to find all the connected regions of the foreground pixels and give distinct labels to each of the different connected regions. Generally, we can classify the algorithm into two types according to its memory access strategy: regular access and irregular access. For the type of irregular access, the propagation of label causes unbounded access to the image. This causes low execution efficiency. For the type of regular access, it usually uses two or more raster scan to circumvent the unbounded access problem. The connected-component labeling method proposed in this thesis uses only one raster scan and can be classified as a regular access type. We use a single-direction linked list to store the line segments in each connected region. To prevent unbounded access to the line segments, three additional pointers are introduced to separate the connected region into three parts. These pointers are then used to merge connected regions and prevent potential unbounded access during the merging process. From experimental results, we can see that the execution time of the algorithm depends mainly on image size and the number of line segments. The effect of region merging is significantly reduced. This proves that the new algorithm can effectively reduce the execution time of connected-component labeling.