Video Sequence Based Image Mosaic

In order to preserve the digitallization of medical film is inevitable. Current medical film scanser are expensive. Video capture is a method of digitallization, but often limited by the ocular view. Image mosaic is a way to broad the ocular view. To explore a method based on video sequence image mosaic to digitize the medical film with low cost and excellent performance is of great expectation.To aqcuire a steable and sequential video sequence, which provide high quality images for the following mosaic, we reform the 2-D flat plotter, add illumination light , adjust the camera etc. According to the discipline of movement of camera, we only execute image mosaic with certain video frames, which reduce the time expense. In order to enhance the accuracy of image mosaic, we intensify the detail of image by executing image graying and gray stretching before image mosaic. Image mosaic includes image registration and image fusion. When calculating the registration parameters, common phase correlation method only calculates the displacement value of two images when they are mostly matched. Multi phase correlation method set the displacement values which is calculated by common phase correlation method as candidate value, then execute registration with several diaplacement value rather than execute registration by the position with max correlation value. In each registration solution, once again we intercept sub images and calculate the displacement value with phase correlation method, then compromise the two displacement result and form the final mosaic parameters. With the mosaic parameters, we should not simply align the images, but adopt certain fusion algorithm to erase the mosaic trace and light diversity of the wrap area between two images.Experiment on 50 cases (more than 2000 images) shows that the system is robust and adaptive, the result is good on the whole.