This article proposes an optical-analog technique for THz image processing based on a polarization-multiplexed metasurface to perform concurrent differentiation and integration. Polarization encoding enables the processor to effectively integrate two optical transfer functions corresponding to differentiation and integration into a single-layer metasurface with fewer hardware constraints due to immature THz device development and less impact by physical barriers such as the diffraction limit. Experimental results demonstrate concurrent differentiation and integration operations on THz images for edge enhancement and denoising, paving the way for multifunctional, high-speed THz image processing with the potential to accelerate data handling and preprocess data for digital neural network algorithms.