Contour Point Signature in Registration Tomography and Magnetic Resonance Images

Resumen

Medical image registration plays a crucial role in several clinical applications, where the alignment of image modalities is essential for accurate diagnosis and effective treatment[1, 4]. To address this registration challenge, we explore an innovative feature-based approach using the Contour Point Signature (CPS) proposed in [2]. Given a contour A, whose reference points are P = {p₁, p₂, . . . , p_N}, p_i ∈ R², the Contour Point Signature relative to point p_i represents the relative distance distribution of contour points, which is a unique signature for each shape. Taking the signatures of all points in P, we obtain a matrix whose ij-entry is a descriptor feature matrix of the shape. The Matrix stores the signatures of contour points, providing a structured representation of the shape based on its contour, capturing the object’s form. [...]