Artificial Intelligence is increasingly vital to high-stakes industries like healthcare and finance, yet these systems often falter when processing the noisy, imperfect data found in real-world environments. While Convolutional Neural Networks (CNNs) are the standard for image classification, Quantum Neural Networks (QNNs) theoretically offer superior noise resilience through quantum properties such as superposition and entanglement. However, existing comparative research is often inconsistent, frequently pitting high-resolution classical models against hardware-limited, low-resolution quantum simulations. This research addresses that disparity by conducting a comparative analysis of error mitigation between QNNs and CNNs under Additive White Gaussian Noise (AWGN). My methodology involves developing three distinct models within TensorFlow Quantum using the MNIST dataset: a standard full-resolution CNN, a hybrid QNN, and a resolution-matched fair CNN to ensure a direct architectural comparison. All models will be trained on clean data and subsequently tested against noise-injected datasets to measure classification accuracy and loss. Through this research, I aim to quantify whether quantum architectures inherently outperform classical ones in signal-to-noise processing. Observing superior robustness in the QNN would validate the practical viability of quantum machine learning for critical applications where data integrity is compromised.