A method for detecting defects in a near-eye display is provided. The method includes the following steps: obtaining a reference image and a DUT image according to a first image and a second image captured by a camera through a Fresnel lens when a display panel respectively displays a test-pattern image and a test-background image; performing a fast Fourier transform on the reference image and the DUT image to obtain a frequency-domain reference image and a frequency-domain DUT image; calculating an average value of pixel values above a predetermined cut-off ratio in a histogram of each first region of interest (ROI) in a filtered frequency-domain reference image as a corresponding threshold; comparing each pixel in the filtered DUT image with the corresponding threshold to generate a determination result; and building a defective-status map of the near-eye display according to the determination results.