A Comprehensive Image Segmentation Approach For Image Registration

Image registration is the technique of transforming different sets of data all in one coordinate system. Automatic image registration remains a real challenge in various fields like computer vision and remote sensing applications. Within this project, a way for automatic image registration through histogram-based image segmentation is proposed. This new approach mainly consists in utilizing the pair of images that should be registered are segmented, according to a relaxation parameter by the histogram modes delineation (a new approach), followed by a regular characterization of one's extracted objects through objects area, ratio between the axis as to the adjust ellipse, perimeter and fractal dimension but a robust statistical based technique of objects matching. The most ideal aim of this proposed methodology is tested to simulated rotation and translation. The very first dataset consists in a photograph along with a rotated and shifted aspect of precisely the same photograph, with different stages of added noise. This lets for the registration of pairs of images with differences in rotation and translation.