Calfhood diseases can detrimentally affect productive performance during early stages of growth as well as overall lifetime performance of dairy cows. An automated, real-time assessment of body weight (BW) could be used to monitor growth curves in dairy calves. The objectives of this study was to predict BW of dairy calves using image features extracted from a 3D camera. Forty pre-weaned Holstein dairy calves with age varying from 2 to 8 weeks, and BW of 57.0 ± 14.7 kg (average ± SD) were used to develop the models. A 3D camera was positioned to capture images from the dorsal area of the calves. Thirty seconds of video was recorded and at least 10 frames were obtained for each calf. The image was segmented and 28 features were automatically extracted, including projected body volume, dorsal area, dorsal length, dorsal extent, dorsal eccentricity, distance from the camera to the floor, width in 11 points along the dorsal area, and height from the camera to the dorsal area in 11 points. These variables were then used to develop 4 models: (1) linear regression (full model: 28 variables; LRFM), (2) linear regression (stepwise backward selection: including dorsal volume, width 2, width 8, width 11, and dorsal extent; LRSTEP), (3) partial least squares (PLS), and (4) artificial neural network (ANN). The developed models were validated in an independent data set composed of 30 images from 5 pre-weaned calves obtained between the 4th and 8th weeks of age (63.8 ± 6.7 kg of BW). The biometric features automatically extracted from the 3D camera images were consistently associated with body weight. All linear models presented a good fit to body weights (LRFM: R² = 0.93, LRSTEP: R² = 0.85, PLS: R² = 0.78). In the validation data set, the PLS model presented the best prediction quality compared with all other models (R² = 0.80, mean bias = 6.0 kg, and RMSEP = 6.5 kg). Results indicate that image analysis can be used as a potential tool for real-time prediction of body weight in pre-weaned calves. Monitoring calf growth development in commercial herds can be used to anticipate potential health problems, and hence guide preventive management practices.