Deconvolution technique has been widely used in fluorescence microscopy to restore fine structures of biological samples. However, conventional deconvolution methods usually achieve little contrast enhancement in dense structures that have the intervals close to the Rayleigh criterion. Herein, we developed a novel deconvolution method, termed virtual single-pixel imaging (v-SPI). Differing from existing deconvolution methods, v-SPI aims to retrieve the less blurred image directly, not the sample distribution which cannot be actually obtained. And the result can be retrieved simply by solving a linear matrix in spatial domain. In addition, the proposed method has no requirement of calibrating parameters of microscope system. Simulation and experimental results demonstrated that the proposed v-SPI method can enhance the contrast of dense structures significantly and acquire a 24% increase in resolution.