Color is an important source of information for: image comparisons, retrieval, and analysis. Unfortunately, color is strongly affected by many factors and a suitable color recovery preprocessing step is necessary to fully exploit the power of chromatic features. This paper describes a new hybrid technique for color recovery, which gives improved performance even in presence of sensible environmental changes. The technique combines one of the most effective retinex algorithm (McCann99) with the gray-world transformation applied at multiple resolutions. A non linear post processing step is also proposed that improves the final color balance and range. Extensive experimental tests have been performed both on synthetic and real images, that confirm the robustness and the enhanced color recovery capability with respect to other on the edge techniques. The proposed method has been used in a real time system for the automatic recognition of those plates that denote vehicles for dangerous-goods transportation. Some results about this last system are reported in the paper.