Since PCA-Based Teeth-Image PersonalIdentification Method [3] and LDA-Based Teeth-ImagePersona Identification Method [2] are not robustagainst reflection and orientation, registered persons inthe database is rejected around 10%. This paperproposes a method to improve the PCA &LDA basedteeth-image personal identification method. In thismethod, the teeth image failed from the matching in thePCA &LDA based system is reconsidered by feedingback the image to eliminate the reflection and therotation problems. The enhanced teeth image is fed backto PCA &LDA process in order to rescue misclassifiedteeth-image. In the experiments, 500 teeth images aretested with 200-teeth database. The results revealed thatof the 10% errors caused by the two problems, 5% arecorrectly identified because of the proposed method.