Speech source localization aims to estimate the relative location of speech sources (speakers) using speech recordings of a microphone array, which is an important function in many applications, e.g., robot audition, video conferencing, smart loudspeaker, etc. In indoor environments, speech signal propagates from speaker to microphones through many paths, including the direct path and the reflection paths of walls. Reflections and reverberation contaminate the direct-path signal in microphone recordings and thus make source localization very difficult. In this chapter, a reverberation-robust localization feature, called direct-path relative transfer function (DP-RTF), is introduced. Propagation paths can be encoded in room impulse responses (RIRs). The basic principle of DP-RTF is to perform blind RIR identification in the short-time Fourier transform domain and then extract the direct-path component as (reverberation-free) localization feature. The definition, estimation and clustering of DP-RTF will be systematically presented, as for various configurations such as noise-free and noisy environments, two and multiple microphones, static and moving sources, single and multiple sources. Experimental examples will also be given to illustrate the properties and capabilities of DP-RTF.