Numerous image processing methods have been proposed to help low vision people, often relied on contrast enhancement algorithms. Their assessment is usually performed by tests on low vision subjects, which are expensive and time consuming. This paper presents a low vision observer model, fully customizable to fit various impaired visual performances, which may be used for early algorithm assessment, and avoiding unnecessary human tests. This model is fitted to visual performances of a subject with degenerative retinal disease, and applied to images processed by two edge enhancement algorithms, allowing to explain their performances in terms of blur reduction and color saturation improvement.