Encoding of visual information in the brain is retinotopic: Neighboring points in the visual field are mapped onto neighboring photoreceptors in the retina, and these neighborhood relations are maintained in the early stages of cortical processing. However, perception is nonretinotopic. First, vision is stable and continuous although the retinal image is in constant flux and frequently interrupted by eye blinks. Second, object parts are perceived relative to the object, rather than in retinal coordinates. For example, the reflector on the wheel of a moving bicycle is perceived to circle, although it follows a cycloidal trajectory across the retina. The precise mechanisms of how, where, and when visual information is transformed from a retinotopic into a nonretinotopic representation, remain largely elusive. While nonretinotopic perception across eye movements is relatively well understood, almost nothing is known about nonretinotopic processing without eye movements. Here, I used psychophysical experiments to uncover the perceptual mechanisms that are at play. My results show that A) the computation of nonretinotopic perception does not require predictability of the stimulus, suggesting that the computations are very fast; B) unconscious processing of retinotopic motion can interfere with the conscious nonretinotopic motion percept; C) the computation of nonretinotopic motion is likely a two-stage process, in which first a reference motion is computed, that is then subtracted from the retinotopic motion; D) nonretinotopic motion perception is intact in schizophrenia patients; E) endogenous attention affects nonretinotopic feature integration primarily by increasing the resolution of the initial stimulus encoding.