Abstract.Image databases obtain the incremental images frequently. However in tradition it is directly compressed to the incremental image, which ignores the redundancy and similarity between images. The general compression way is not satisfied. With an increasing number of compression method proposed, we proposed a method called incremental image compressing method aiming to this question, which is to make the best of the redundancy and similarity between image database and incremental image. Firstly, we tried to classify the incremental image, then seek the most similar image in the class as the reference image, and encode the incremental image using the redundancy and similarity. Simulations test shows the compression results are obviously improved, so the proposed method is superior to traditional compression methods.