Innovative 3D Reconstruction Method based on Patch Based Technique using Neural Network

In this paper, the innovative 3D reconstruction method of human skull, based on the patch based technique using Convolutional Neural Network (CNN) is proposed. Firstly, the image filtering to reduce the influence from noise to mode detection was applied. Next, the filtered image was split into segments using proposed CNN (accurate image segmentation). The proposed CNN for image segmentation consists of several layers. Each layer occupies a multi-dimensional array of numbers. The proposed network composed of layers that transform an input image from the original pixel values to the final layer score (layer by layer). The small segments were merged together to the most similar adjacent segments. Finally, the patch based technique (patch extraction) for 3D reconstruction of the skull was used. The experimental results using biomedical data (CT data) indicate the effectiveness of the proposed reconstruction approach.