A Research On The Inter-Slice Interpolation Methods For Medical Images

Inter-slice interpolation is often a must prior to 3D reconstruction of medical images, which is largely necessary from the generation to the post-processing of images. Generally speaking, due to the through-plane resolution of medical images is much lower than their in-plane resolution, the original anisotropic data can be transformed into isotropic discrete data by way of the inter-slice interpolation of images. Although there are many image interpolation methods available in the published literature, most of them are not able to give consideration to both gray levels and objective shape variation of the images. Moreover, the calculation process is relatively complicated. Therefore, it is imperative and important to find a highly efficient and widely applicable interpolation method for the 3D reconstruction of medical images. The interpolation algorithms were classified into three groups: gray level-based methods, object-based methods, and wavelet-based methods in current work.The pre-processing methods of medical images include image denoising, histogram equalization, image sharpening etc. After the discourse on the ideal sinc interpolation function, the gray level-based interpolation was then introduced which includes the linear interpolation, the B-spline interpolation, the cubic interpolation, and the Lagrange polynomial interpolation. And the object-based interpolation method was discussed as well. All these methods were then evaluated by using the same image dataset in this part of the work.The wavelet-based interpolation method was emphatically introduced in this study. The method combining wavelet transform with polynomials was proposed to deal with medical images at different scales because of the wavelet’s characteristic of multi-resolution. Lagrange polynomial and Gaussian interpolator were chosen as the polynomials. Firstly, the original images were decomposed using the wavelet analysis to attain the positions of the wavelet coefficients which belong to the edges. Then, the polynomial was applied to interpolate the intensities and positions between the corresponding wavelet coefficients of the cross-sectional images. Finally, the resultant interpolation image was achieved by the inverse wavelet transform. Compared to the linear interpolation and the Cubic interpolation methods, the proposed novel method can assure better image quality and reduce the calculation error. The interpolated images can be used to efficiently perform the 3D reconstruction for the object tissues of medical images.In the end, the Marching cubes surface reconstruction algorithm was used to render the 3D image dataset produced by both the linear and the proposed image interpolation methods for comparison.