An optimized CT image retrieval method

In recent years, the medical image retrieval play an important role in the field of medical diagnosis. The primary goal is to retrieve accurate matching images from the database. In order to achieve this goal, quite a few methods were used in the past years and some of them can get realistic results. However, after some deep and further experiments, we found that some classic feature descriptors or methods are not effective for all images. Especially for some special images, the results can't meet our expectation. In other words, the previous methods are not universal. After some further research, we found that the majority of CT images contain obvious edge or shape details, other feature are not obvious. For this problem, we combine some previous methods and improve them. In the process, Gabor transform, LBP(Local binary pattern) and Hu moment are combined for solving this problem. At the same time, some optimized methods are also used. At last, we do some groups of experiments using our new method. Enough groups of experiments show our method's superiority. Although our method is not complicated, but it's effective for retrieving the target images. Especially for some special images, our results are also as realistic as other common images.