The stored microscopic images of locust slices often turn out to be folded owning to improper operations during experiments. As these images cannot be reproduced in practice, it is necessary to restore the images of the folded slices to accurately reflect the original biological tissues of the locusts. The nearest unit matching method was used to restore the damaged images of locust slices. A sequence of images of slices was first magnified by a factor of two to extract the feature points from the folded slices and the reference slices using SIFT (Scale-invariant feature transform) algorithm. The initial matching pairs were determined using the k-d tree strategy and the incorrect matching pairs were eliminated using RANSAC (Random sample consensus) algorithm. The folded area and the rest of the image were then divided into sub-cells respectively. By comparing the number of matched pairs near the folded sub-cell, the matching sub-cells in the image were determined. The space mapping model was established between matching pairs associated with the nearest slice sub-cell of each folded sub-cell by using the least square method. The matching blocks within the reference slice of the folded sub-cell were selected based on the space mapping model. The method was tested by using different locust slices compared with the traditional image restoration methods. Then the common image restoration evaluating indicators were used, such as the peak signal to noise ratio and the mean square error to compare and evaluate the image restoration results. The experimental results showed that the method yielded more feature points and the established space mapping model could better adapt to various veins of the slices with accurate segmentation and restoration of images of damaged slices.