Contactless breath sensing using the radar technique has emerged as a promising solution due to its user-friendly and privacy preservation advantages. However, in real-life situations, diverse body postures lead to complex signal propagation around the body, which complicates breathing motion extraction. Consequently, the effectiveness of radar-based breath sensing can be compromised in everyday scenarios due to varying body postures, highlighting a continuous challenge. In this paper, we propose exploring the entire body reflections and coherently utilizing spatially distributed breath signals with imaging radar for robust breath sensing across various postures. Specifically, we first implement a novel hybrid breath imaging algorithm along with a millimeter wave imaging radar for spatially detailed imaging of breath-like signals from coarse to fine grids. Then, we coherently accumulate breath signals to counteract interference and noise arising from different postures. The experimental results indicate that the proposed method achieves the Mean Absolute Error (MAE) of 0.38 breaths per minute (BPM). Compared to the baseline methods, this approach can reduce the MAE of breath rate estimation by up to 47% and reduce the variance of MAE by up to 75%. This highlights the system’s robustness and effectiveness in handling various posture scenarios.