The document presents a novel approach for single-image depth estimation using Fourier domain analysis and a ResNet-based convolutional neural network (CNN). It introduces a depth balanced loss function (DBE) to mitigate depth bias in training and combines depth map candidates from various cropping ratios through Fourier analysis, enhancing accuracy in depth estimation. The proposed method shows promise for applications in object localization and planetary exploration by providing more reliable depth information from single images.