Abstract:

In order to fuse efficiently the multi-channel information of the target and improve the performance of the filter, a new class of k-mask based hyper-connectivity was proposed. The connectivity mask image and the reconstruct mask image were obtained from the different channel informations and restraint conditions. According to the area information by which the connectivity mask image intersects the reconstruct mask image, the connected components in the reconstruct mask image were relabeled to form the new connection space by selection and aggregation. From the definition of the hyper-connectivity class,  k-mask based hyper-connected open operator and reconstruct operator were designed and applied in labeling the character region and extracting the major path of the remote sensing image. Experiment results illustrated that the filter performance of the traditional open operator and reconstruction operator was improved significantly by flexible selection of the connectivity mask image and fusion of the image multi-gradient information, color information and multi-channel information.

Key words: computer application, connected operator, hyper-connectivity, reconstruction operator, image segmentation