This paper proposes a k-reciprocal encoding method for re-ranking in person re-identification (re-ID). The key idea is that if a gallery image is similar to the probe in the k-reciprocal nearest neighbors, it is more likely to be a true match. The method calculates a k-reciprocal feature by encoding the k-reciprocal nearest neighbors into a single vector, which is used for re-ranking under the Jaccard distance. The final distance is computed as the combination of the original distance and the Jaccard distance. The method does not require any human interaction or labeled data, making it applicable to large-scale datasets. Experiments on the Market-1501, CUHK03, MARS, and PRW datasets show that the method improves re-ID performance, achieving state-of-the-art accuracy on Market-1501 in both rank-1 and mAP. The method also includes a local query expansion approach to further improve re-ID performance. The final distance is calculated as a weighted aggregation of the original distance and the Jaccard distance. The proposed method is fully automatic and unsupervised, and can be applied to any person re-ID ranking result.This paper proposes a k-reciprocal encoding method for re-ranking in person re-identification (re-ID). The key idea is that if a gallery image is similar to the probe in the k-reciprocal nearest neighbors, it is more likely to be a true match. The method calculates a k-reciprocal feature by encoding the k-reciprocal nearest neighbors into a single vector, which is used for re-ranking under the Jaccard distance. The final distance is computed as the combination of the original distance and the Jaccard distance. The method does not require any human interaction or labeled data, making it applicable to large-scale datasets. Experiments on the Market-1501, CUHK03, MARS, and PRW datasets show that the method improves re-ID performance, achieving state-of-the-art accuracy on Market-1501 in both rank-1 and mAP. The method also includes a local query expansion approach to further improve re-ID performance. The final distance is calculated as a weighted aggregation of the original distance and the Jaccard distance. The proposed method is fully automatic and unsupervised, and can be applied to any person re-ID ranking result.