Akademik Veri Yönetim Sistemi
23nd Signal Processing and Communications Applications Conference (SIU), Malatya, Türkiye, 16 - 19 Mayıs 2015, ss.859-862
In this study, a method is presented that objects in gray-level input images are normalized based on the affine transformation. The process of normalization can be expressed as removal of geometric deformations in the image such as translation, scaling, rotation, and skew. In the method, first, the input image is thresholded and then a binary image is generated. In the next step, objects in the binary image are labeled. In the last process step of the method, the labeled objects are normalized. In the normalization process, positive eigenvalues and related eigenvectors involving the covariance matrix belonging to the labeled binary image are taken into account and a tensor approach is also used. With the proposed method, even if the input image has symmetrical objects, the objects can be normalized successfully. The obtained visual results prove the performance of the method.