Estimation of body parameters, height and body weight, is important for biological profiling in forensic investigations. This study aimed to develop a deep learning model for automated height and weight estimation using postmortem computed tomography (PMCT) scout view images acquired in forensic practice. A retrospective analysis was conducted in 3,124 cases that underwent autopsy and PMCT. Scout view images were preprocessed with geometric correction based on PMCT table height. Three CT window settings were fused into RGB channels, enhancing anatomical and environmental information. A convolutional neural network based on a pre-trained VGG16 architecture was trained to simultaneously estimate height and weight. Model performance was evaluated on a test dataset (n = 624) using mean absolute error (MAE), mean absolute percentage error (MAPE), correlation analysis, and Bland-Altman analysis. The multi-window RGB fusion model achieved MAE values of 2.39 cm for height and 1.83 kg for weight, outperforming the single-window model. Predicted and reference values were strongly correlated. Bland-Altman analysis demonstrated narrow limits of agreement. Deep learning applied to PMCT scout view images enables automatic estimation of height and weight with high accuracy. The proposed multi-window RGB fusion with geometric correction offers a practical tool for body parameter estimation in forensic investigations.