EFFICIENT TILE BASED REGION MERGING REMOTE SENSING IMAGE SEGMENTATION

Veena V. Patil, S. G. Shinde

Abstract: The remote sensing technologies commonly capture the high resolution images of Earth and transmit them to central process department for further analysis via electromagnetic sensors. Such remote sensing images are stored in digital forms in computers in which further processing is performed for the interpretation purpose. However, processing of high resolution remote sensing large images on computers is challenging due to the large sizes of remote sensing images. For common computers, it is difficult to hold big sized remote sensing datasets for longer time. To solve this problem, segmentation is best solution to save resources for further processing. The remote sensing segmentation method produces the objects of interest from the scene. There is various image segmentation methods reported in recent pasts, but scalability, efficiency, and accuracy is main challenges. In this paper, we proposed tile based image segmentation algorithm. We designed tile-based framework based on region merging algorithm for large images segmentation with identification outcomes. We exploited the stability margin concept for the tile. Additionally, to smooth and fast the process we proposed the robust filtering technique which helps to suppress noise as well as complex data. The filtering approach achieves the improved segmentation results within less or similar processing time. The simulation results shows that proposed method produces more robust results as compared to existing tile based and without tile based methods

Keywords: Filtering, Image processing, Image Segmentation, Region-Merging, Tilting, Scalability

DOI: https://doi.org/10.15623/ijret.2018.0705003