Detecting sharp illumination variations such as shadow boundaries is an important problem for radiosity methods. Such illumination variations are captured using a nonuniform mesh that refines the areas exhibiting high illumination gradients. Nonuniform meshing techniques like discontinuity meshing often rely on shadow casting, so their application is computationally expensive. This paper presents a sharp shadow detection technique for radiosity tools based on the progressive refinement method. The presented technique offers good results (especially for capturing sharp shadows cast by small ’detail’ objects), is very simple to implement, has negligible time and space requirements and integrates well with other adaptive subdivision strategies in a radiosity tool based on progressive refinement.