Abstract
To speed up the processing time, we decentralized the processing tasks to generate new view. In proposed methods, each sensor node processes part of corresponding search algorithm based on MSE minimization to generate the interpolated image with local communication between sensor nodes. In conventional method, all the processing tasks to generate a new view were centralized in central node. We proposed two methods, which are Fully Image Shared Decentralized Processing (FIS-DP), and Partially Image Shared Decentralized Processing (PIS-DP). In FIS-DP, full image transformation should be done between neighbor camera sensor nodes, but the correspondence check is performed in different block intervals. In PIS-DP, the correspondence check is performed in part of images in each node by sharing part of captured images. The experimental results of FIS-DP, and PIS-DP show that we can increase the processing speed of the whole network to generate a new view in compare with Centralized Processing (CP) method. The PIS-DP method is recommended according to its better processing speed and communication efficiency in compare with other methods (FIS-DP, and CP). Using PIS-DP, will let us to perform real-time free view generation in camera sensor network.
