This paper proposes an efficient algorithm for personal identification with biometric images. In identification based on image comparison, the number of comparisons is an important factor to estimate the total processing time in addition to the processing time of a single comparison. Maeda et al. proposed an identification algorithm that reduces the number of comparisons from the linear search algorithm, however the processing time of each comparison is proportional to the number of registered images. The algorithm in this paper is an improvement of the algorithm by Maeda et al. with constant-time image comparisons. This paper evaluates the algorithms in terms of the processing time and the accuracy with practical palmprint images, and proves that the novel algorithm can reduce the number of image comparisons from the linear search algorithm as the algorithm by Maeda et al. without loss of the accuracy.