The high prevalence of chronic wounds and their consequences for people’s quality of life makes wound treatment a highly relevant topic in the context of healthcare. One vital aspect of monitoring concerns tracking wound size evolution, which guides healthcare professionals during diagnosis and serves as a key predictor of treatment efficacy. This work proposes an automatic image segmentation and measurement framework for chronic wounds using deep learning and computer vision techniques. The wound segmentation task involved exploring three prominent segmentation models: a popular convolutional neural network (DeepLabV3+), a cutting-edge transformer approach (SegFormer) and a visual foundation model (MedSAM). Traditional computer vision techniques were further applied to infer the open wound’s width, length and area in real-world units during the wound measurement task. Separate studies were performed to assess each task’s performance and a final assessment of the complete framework that couples a wound and reference marker detection model with the developed segmentation and measurement approach. For the automatic wound segmentation, MedSAM achieved the best performance with Dice scores of 88.14% and 92.25%, applied on public AZH FU and private datasets, respectively. In the wound measurement task, the area estimation achieved a mean relative error of 5.36% for the private dataset. Concerning the overall pipeline results, MedSAM experienced a decline in performance and SegFormer emerged as the best segmentation model, achieving a Dice score of 91.55% and a 16.7% mean relative error for the area estimation (which surpasses the literature results) in the private dataset, demonstrating its applicability in clinical practice.