Image segmentation plays a pivotal role in computer vision and image processing, and active contour models (ACMs) emerge as one of the most prevalent and widely adopted algorithms for achieving image segmentation. One notable advantage of ACMs is their ability to produce accurate segmentations for image objects with intricate structures, varying locations, and sizes without the need for training samples. In recent advancements, ACMs have achieved substantial enhancements in accuracy, robust initialization, and optimization speed by integrating edge information within the level set framework. In this study, we introduced a new ACM based on fractional order derivative that addresses multiplicative noise and non-uniform intensity distribution in the segmentation task by incorporating two crucial terms: image segmentation and despeckling, with the former mitigating noise effects and the latter refining initial contours for accurate boundary delineation. The fractional term is discretized using the Grünwald–Letnikov (G-L) derivative, while the forward time and central space scheme is applied to the integer order term. In extensive numerical experiments on noisy and heterogeneous natural and medical images, the present model outperforms current ACMs in terms of performance, as assessed through various evaluation metrics.