Summary: Image segmentation, under the broader area of computer vision and digital image processing, facilitates comprehending a digital image by partitioning it into small and meaningful segments, regions or objects. Researchers have come up with numerous solutions, which have produced promising results. However, these results are compromised when applied to global segmentation, where all the objects/features are needed to be segmented for inhomogeneous and severe noisy images. The reasons are many including (i) the use of region-based active contour method, especially when its noise distribution is unknown; (ii) lack of robustness to handle images corrupted by some multiplicative and additive noise; (iii) inability to handle convex images at local minima. To fill this gap in the literature and come up with a more effective solution, this paper proposes a global segmentation model that implements a local image difference technique combined with average convolution local factor based on spatial distances and intensity differences in the local region. The proposed model has been tested and evaluated on outdoor, synthetic, and medical images, where it has outperformed the latest baselines. The model is capable of tackling efficiently noise/outliers and coping with intensity inhomogeneity issues.