| Gastrointestinal mucosal disease and thus cause further digestive tract cancer is one of the biggest killers of health. Traditionally, doctors used fiber endoscope to diagnose digestive tract diseases, but this diagnostic approach was inconvenience and painful, so many patients gave up the diagnosis. To resolve this problem, a capsule endoscope has been invented to alleviate the suffering of patients. But doctors can not accurately diagnose the lesion area because of the huge amount of capsule endoscope images, and the movement blurred of images. Hence the capsule endoscope images should be deblurred to improve the accuracy of the lesion area diagnosing.Image deblurring is a process of deconvolution using blurred image and blur kernel. Deblurring of capsule endoscope images is a blind deconvolution problem because the blur kernel is unknown. So capsule endoscope image deblurring can be divided to two steps. First, solve the Maximum a Posteriori problem to estimate the blur kernel. Second, use partial differential equation to restore the blur image. The traditional deblurring algorithm will produce ringing artifacts during the deblurring and influence the effect of the deblurring. And also the traditional algorithm can't retain the texture of the latent unblurred image. To resolve these problems, first, using a local smoothness prior to estimate the blur kernel can reduce ringing artifacts during the image deblurring. Second, the total variation model in Partial Differential Equations has the different diffusivity of the gradient direction and the parallel direction. Using the total variation model in regular term to deblur the capsule endoscope images can retain the main textures of images.The proposed algorithm was simulated using Matlab, 563 blur small intestine images from a database were selecte to verify the proposed algorithm, about 520 images had good effects of deblurring, and the deblurred images were proved effective to diagnose diseases. |
PDF Full Text Request