An interlayer interpolation method for MRI images based on matching of corresponding points in local regions
Objective Interlayer interpolation in medical images is very important for improving image resolution. Some traditional algorithms based on grayscale interpolation often lead to the loss of tissue details or blurred edges when dealing with regions with complex structures and large grayscale variations between layers. The aim of this study is to establish a local region corresponding point matching interpolation algorithm,so as to effectively improve the quality of the interpolated images. Methods In this study,a local region corresponding point matching algorithm is constructed. The error image is constructed by comparing the image obtained by linear interpolation with the original scanned image at the same location,and high error regions are identified. Linear interpolation is used for the non-high error regions. The corresponding point matching method is used for the high error regions. It first determines the best parameters with the minimum error of the matching error function of the point pairs used for interpolation based on the local information of the region,and then finds the best matching point pair on both sides of the image for each interpolated point in the region for the grayscale calculation of the interpolated points. This method is applied to the interlayer interpolation of axial,sagittal and coronal scanning images,and the mean-square error between the interpolated image and the original scanning image is used as the criterion to evaluate the quality of the interpolation. Results Compared with the interpolation results of other algorithms,the local region corresponding point matching algorithm can significantly reduce the interpolation error,especially in sagittal image interpolation. Compared with the results of the overall correspondence point matching algorithm,the average error can be reduced by 19％,indicating that the quality of the interpolated image in the original high error region is effectively improved. Conclusions The local region corresponding point matching interpolation algorithm established in this study can effectively solve the shortcomings of previous algorithms and significantly improve the accuracy and quality of interlayer interpolation of images.
magnetic resonance imageinterpolation of medical image slicescorresponding point matching interpolationlocal region method
万方数据
维普
