Lighting and image-processing techniques for automatic fuel-number recognition in boiling water reactors (BWRs) are described. High-contrast images of fuel numbers are obtained for various types of fuel subassemblies by introducing shallow-angle lighting. Moreover, high-quality binary-valued images can be obtained by multidirectional lighting and character reconstruction techniques. These techniques are experimentally confirmed using test subassemblies. The multidirectional lighting is verified to be superior to unidirectional lighting in terms of its recognition rate and its applicability to an automatic fuel-number recognition system for BWRs.