Thyroid associated ophthalmopathy (TAO) is one of the most common orbital disease, it can be easily detected by the human eye in the late onset due to obvious changes in extraocular muscles. But in the early stage, it is not easy to be distinguish by doctors with eye because of subtle changes in extraocular muscles, so it is a good way to use the computer’s ability to assist doctors to pre-diagnosis of TAO for follow-up treatment. In this paper, according to the routine diagnosis process of doctors, a comprehensive detection system network is proposed. The network consists of three different convolutional neural subnetwork, corresponding to three bitmaps of eye CT images .Finally, the output of three subnetwork are combined to generate final diagnostic result by the majority vote. Through the experiment, the detection system, whose recognition rata is 94.87%, has a good ability to identify the characteristics of TAO, can assist the doctor in the early diagnosis of TAO in a certain extent, so as to help early patients get timely treatment.
This paper expounds method of the average weighted fusion, image pyramid fusion, the wavelet transform and apply these methods on the fusion of multiple exposures nighttime images. Through calculating information entropy and cross entropy of fusion images, we can evaluate the effect of different fusion. Experiments showed that Laplacian pyramid image fusion algorithm is suitable for processing nighttime images fusion, it can reduce the halo while preserving image details.