In recent years, localization for capsule endoscopy applications using Ultra-Wideband (UWB) technology has become an attractive field of investigation due to its potential benefits for patients. Literature concerning performance analysis of RF-based localization techniques for in-body applications at UWB frequencies is very limited. Available studies mainly rely on Finite Difference Time Domain (FDTD) simulations, using digital human models and on experimental measurements by means of homogeneous phantoms. Nevertheless, no realistic analysis based on multilayer phantom measurements or through in vivo experiment has been reported yet. This paper investigates the performance of RSS-based approach for two-dimensional (2D) and three-dimensional localization (3D) in the UWB frequency band. For 2D localization, experimental laboratory measurements using a two-layer phantom based setup have been conducted. For 3D localization, data from a recently conducted in vivo experiment have been used. Localization accuracy using path loss (PL) models, under ideal and not ideal channel estimation assumptions, are compared. Results show that under not ideal channel assumption the relative localization error slightly increases for the 2D case but not for the in vivo 3D case. Impact of receivers selection on the localization accuracy has also been investigated for both, 2D and 3D, cases. Index Terms-Wireless capsule endoscopy (WCE), Ultra-Wideband (UWB), in-body localization, heterogeneous phantom, in vivo measurements.