High-resolution image acquisition is the premise and foundation of automatic pattern recognition. We study the multi-focus image acquisition problem of stereoscopic pests, taking the rice field insect as the object. Harris corner detection and image entropy are taken as the image quality detection standard. The influence of step displacement in image acquisition on the resolution of image fusion is analyzed with image fusion method based on wavelet transform for different scales of rice pest images. According to the experimental comparison, we obtain the best image acquisition and image fusion strategy and the relation of magnification and optimum step displacement. The experimental results show that this strategy is reliable to acquire the extended-depth-of-field image for stereoscopic pests, which provides an effective measure for establishing a high-quality image database of rice pest samples.