We present a new feature-based design concept for filter design in which a filter is designed specifically to image a known feature with dimensions lower than the optical resolution of the system. Unlike the conventional filter design based on focal spot engineering, we use the complete response of the microscope to form a resolution factor and consider minimizing the resolution factor as the design goal. We consider three design goals (i.e., resolution factors) based on the system response and show that a feature matching approach is more suitable in practice. It is shown that a three-bar pattern and grid of square dots of critical dimension, 0.08λ, can be resolved using a simple two-layer feature-based filter designed for radially polarized illumination in aplanatic solid immersion lens microscopy. An example that requires a more complex filter is also shown. Applicability of the concept of feature-based filter design in diverse microscopy scenarios is discussed as well.