This paper investigates the effect of reducing the number of microphones on noise suppression performance within an instantaneous-estimation‑based spectral subtraction (ISSS) framework employing spectral image inpainting. Previous studies have shown that ISSS achieves effective noise suppression by estimating the noise component through inpainting of the spatial direct‑current (SDC) bin in a two‑dimensional spatiotemporal spectrum using an eight‑channel microphone array. With practical hardware implementation in mind, we evaluate whether comparable SDC estimation performance can be achieved with fewer microphones. We perform a comparative evaluation using four‑ and eight‑channel linear microphone arrays with comparable apertures. Experiments under single-noise-source conditions show that the four-channel configuration achieves noise suppression performance comparable to that of the eight-channel configuration, while occasionally exhibiting improved performance under specific conditions. These results indicate that noise estimation by inpainting the SDC component depends more strongly on spatiotemporal structure than on spatial sampling density, offering insights into compact array design.