Photon scattering-induced wavefront aberrations impose significant limitations on optical imaging within scattering media, particularly in environments characterized by diffusion. Scattering imaging techniques leveraging memory effects offer a promising avenue for imaging under the regime of multiple scattering. Nonetheless, the presence of interference phase traps inherent to scattered light often results in failures and distortion. We propose an algorithm for computational imaging to address the challenges. By employing gradient-based calculations to delineate the phase distribution of photon ensembles amidst diffuse scattering, we preemptively assess the convergence or divergence of the state of the retrieval algorithm prior to recovery. This proactive strategy serves to circumvent disruptions caused by interference traps, expediting the identification of the accurate state from a myriad of stochastic photon projections, thereby facilitating rapid and high-fidelity scattering retrieval. Through rigorous experimentation involving dozen-group targets and each group with 100 repeated trials, we quantitatively evaluate the proposed method. Our research demonstrates a substantial enhancement in success rates, approximately 3 times higher than those without interference mitigation, while achieving a reduction in computational time, down to 0.2 of the original. This approach introduces a novel scattering imaging technique, 'achieving more with less,' offering technical support for dynamic video imaging in intricate scattering environments such as biological tissues and atmospheric conditions.