The present invention achieves a reduction in the size of a device by acquiring a plurality of images in different position without shifting microlenses (MLs) of an one imaging element while suppressing shading and crosstalk. Provided is an endoscope image-acquisition device including: an objective optical system that forms two subject images in parallel； and an imaging element that has a plurality of MLs arrayed at an entrance side from the objective optical system and in which pixels are allocated to the respective MLs, wherein the centers of light-sensitive portions at the pixels are displaced with respect to the optical axes of the MLs such that displacements therebetween are gradually increased from the center portion of the imaging element toward peripheral portions thereof； the position of an exit pupil of the objective optical system is closer to an object than an imaging position of the objective optical system is； and conditional expression (1) is satisfied.0.5≦(θH−θD)/α≦3 (1)Here, α is an angle formed by a chief ray at a horizontal maximum image height and an optical axis, θH is a chief-ray correction amount (angle) for the ML at the horizontal maximum image height (H) from the center portion of the imaging element, and θD is a chief-ray correction amount (angle) for the ML at a distance D from the center portion of the imaging element. The distance D is the distance between the center portion of the imaging element and the point of intersection between α line drawn toward the center portion of the imaging element at a degrees from the position of the exit pupil of the objective optical system and the imaging element.
