Beamlet-based Multi-scale Image Analysis And Its Applications In The Ct Image Processing

With IT's rapid development and hardware and software's advancement , image processing has been hot and important in the field of signal processing, and it includes image edge detection, segmentation, feature analysis, match, fusion, enhancement, restoration and etc. Such applications are supported by a series of mathematical methods, ranging from traditional Fourier Transform (FT), Discrete Cosine Transform (DCT), Radon Transform and Wavelet Transform (WT) to novel Curvelet Transform and Beamlet Transform. Each novel transform mentioned above, including Beamlet Transform, introduces new algorithm and successful application.The essential difference among those transforms is choice for transform base. This paper reviews some ordinary transforms, and then analyzes the principium of Beamlet Transform. The following part is focused on improved algorithm advantageous over traditional ones with low arithmetic efficiency. For instance, we preserve max and min values in a small scale for an image with less useful information can be presented sufficiently by them. As for feature extraction from polluted image, we define a length threshold to eliminate coefficients corresponding to short beam while preserving useful information. This improved method achieves effective result in line feature extraction for binary image and edge detection for CT image.