Medium depolarization imaging by catheter-based polarization-sensitive optical coherence tomography (PS-OCT) can provide valuable insight into significant features of lipid, macrophages, and cholesterol crystals in atherosclerotic vulnerable plaques. In this paper, we demonstrate a method to achieve an accurate estimation of the medium depolarization index (EMDI) with noise immunity in catheter-based PS-OCT. EMDI is calculated by an iterative approximation based on Lu−Chipman matrix decomposition and Frobenius norm judgment of incoherent averaging of Mueller matrices. Monte Carlo simulation results verify that the medium depolarization measurement by EMDI is 3.3 times more accurate compared with those of the depolarization index (DI) and degree of polarization uniformity (DOPU). In experiments, we design a microsphere suspension with various concentrations and measure EMDI under different additive noise. Consistently, the measurement accuracy by EMDI is increased 2.85 times compared to those by DI and DOPU. For vascular plaques detection, we use protein and cholesterol gel as plaque phantoms. Based on PS-OCT images of plaque phantom in vitro and in ex vivo porcine coronary artery, the recognition rate of plaque by EMDI is 2.99 to 4.65 times higher than those by DI and DOPU evaluated by spatial response of the Laplacian operator (SRLO).