Similar image retrieval only using one image

In this paper, we present one method that can retrieve the similar images only using one image. In recent years, we have used different ways to achieve image retrieval. However, if we use the unsupervised method to achieve image retrieval, the accuracy of image retrieval is reduced obviously. Even if we use the supervised method, the computing time is too long because we need to learn quite a few learning instances. We use best feature descriptor selected, image optimization, deep learning technique to retrieve the target images that is similar to original image. On the one hand, we can see that the method makes full use of image information and select the most effective feature descriptors. On the other hand, we increase the accuracy through optimizing the target images and deep learning technique, so that it is convenient for us to extract more effective information directly. At last, we set up one big database that contains images from different categories. The images are as more complicated as possible. The experimental results show that our method not only can save the computing resource but also can keep the accuracy.