Segmenting images into different multiple parts or regions using AI models has a range of important applications, from autonomous vehicles to healthcare.

Active contour models (ACMs) have been widely used in image segmentation to segment objects. However, when it comes to segmenting images with a high level of variability, most current frameworks do not perform well, which can make it difficult to achieve the desired results.

To address this issue, this DCU research collaboration proposes a new decision-making model, which involves using enhanced local direction pattern (ELDP) and local directional number pattern (LDNP) texture descriptors to create an encoded-texture ACM. The principal component analysis (PCA) is then used to optimize the two encoded images and reduce the correlations before they are fused.

This approach enables the development of a model capable of directly building complex decision boundaries. The experimental results show that the proposed encoded-texture ACM outperforms many recent frameworks in terms of robustness and accuracy for segmenting images with intensity inhomogeneity, fuzzy boundaries, and noise.