IMPROVE IMAGE REGISTRATION JEFFREY?S DIVERGENCE METHOD FOR INSUFFICIENT OVERLAP AREA USING KMEANS++ IN REMOTE SENSED IMAGES

摘要

In remote sensing, lacking sufficient overlap area is a common problem for image registration. To address this issue, Jeffrey?s divergence intensity-based registration technique was developed. This technique is not robust enough when dealing with multimodal images because it influences by the amount of variance in the data, so it may fail to find the optimal registration. Image segmentation can help to reduce the difference between the multimodal images while keep the salient features. kmeans++ was adopted for image segmentation because of it simple and efficient. This segmentation help Jeffrey?s divergence to be more robust with local intensity variation and get the optimal registration even with smaller overlap area. Comprehensive results were conduct to shows the impact of the proposed method to get a better result to compare with the state-of-the-art methods, Jeffrey?s divergence (JD) and mutual information (MI).