The Research Of Image Connected Component Labeling Based On FPGA

With the development of technology and improved quality of life, real-time imageprocessing techniques have been applied more widely. In the process of digital imageprocessing, connected component labeling typically is one of the most importantprocessing stages. It is a basic algorithm in the field of target recognition and tracking.The correctness and labeling speed affects the performance of the entire system directly.In the previous studies, the connected component labeling of static imagesoccupied a large proportion in this field. However, the topic of this dissertation is howto labeling a image when it is in a continuous stream of a video. To solve this problem, anew algorithm is proposed. The feature of video stream is the existence of blanking timebetween two frames. During blanking time, the preliminary label results will beprocessed. When active pixels is coming, foreground pixels in this frame will be labeledusing different numbers. During blanking time, the preliminary information collectedbefore blanking time will be processed. This approach avoids a reasonable accumulationof data in the video stream. By this way, every foreground pixel in the video stream canbe processed during one pixel clock time, meanwhile, the storage capacity of RAMswhich used to buffer images is reduced.In this research, digital image processing is performed on the FPGA. In the end ofthis article, not only the correctness of the proposed approach is verified, but also theperformance of the algorithm is analyzed.