Spiking Neural P System with weight model of majority voting technique for reliable interactive image segmentation

Interactive image segmentation is a method for precisely segmenting of the object from background using information entered by the user. However, most interactive segmentation techniques are sensitive to the location and the number of seed points. To obtain a satisfactory result, the user should repeat the segmentation process over and over, and also based on employed technique, it may work well in some limited conditions and applications. To overcome these limitations and enhance the robustness of interactive image segmentation algorithm, this paper proposes a parallel fusion model using the majority voting technique, which not only is more reliable than existing methods, but also requires less user interaction. To this end, at first the input image is segmented by several segmentation methods independently. Then the obtained results are combined using majority voting technique to extract final segmentation result. To reduce the computational overhead of the proposed scheme, a spiking neural-like P system model for parallel implementation of majority voting technique is also proposed. The proposed model has been evaluated and compared with state-of-the-art methods using different metrics, and the obtained results show its efficiency compared to other methods.

Automated summary

Interactive image segmentation is a method that uses user input to accurately segment objects from the background. Current techniques are sensitive to the location and number of seed points, requiring users to repeat the process multiple times. This paper proposes a parallel fusion model using majority voting technique, which is more reliable and requires less user interaction. Evaluation and comparison with state-of-the-art methods demonstrate the efficiency of the proposed model.