| The digital image processing system is a focus research field in modern information processing. The early days of digital image processing is mainly concentrated in the lower level, including image filtering, image enhancement, image conversion, image storage and image transportation. The main purpose is to facilitate for people to read and understand picture. With the development of computer performance and pattern recognition technology, image processing have been developed to computer vision, artificial intelligence, so that the machine has a capacity to read and understand image, which can substitute for the human beings to engage in intelligent detection, dynamic tracking, or provide support information for human decision.With the development of digital image processing technology, computer technology and storage technology, people have real-time access to the high-resolution image information and to process them, but for advanced digital image processing system, the time performance is an important for a well-designed digital image processing system established not only to complete the function, but also to meet a certain time characteristics, in order to meet the actual demand. Digital image edge detection is a basic field of digital image processing, the edge information is extracted from the original digital image, thus converting the large amount of image information to a more streamlined binary image. While streamlining the amount of data, the edge map image is able to save most of information of the structure, position and the relationship between objects in the original image. Inputting the edge mapping image to advanced digital image processing system can greatly reduce the requirements of the entire system for the storage and computing, enhancing the overall performance of the system.Time-frequency analysis is a powerful tool for modern digital signal processing, which overcomes the drawbacks of the Fourier analysis of losing time-domain information and gives the expression of signal in the time-frequency domain,which makes it possible to observed signal’s local frequency characteristics. They has been most widely used in modern signal processing, in particular non-stationary signal processing. Wavelet transform and S-transform are two time-frequency analysis methods, and have the capacity of multi-scale analysis. Wavelet analysis is available for use with a variety of wavelet bases, giving flexible options according to the different characteristics of the signal or application requirements. The S-transform amends phase information in the wavelet transform and can truly reflect the complete information of the signal in the frequency domain. This thesis mainly investigates the use of time frequency analysis methods in digital image edge detection algorithm.As a time-frequency analysis method, S-transform has ability to detect the singularities and irregular structures in signals, and it is feasible to use S-transform to detect the edge of the digital image. By performing S-transform on the digital image in both the horizontal and vertical direction, and then using the non-maxima filtering and threshold filtering edge, we get the edge mapping. Then a scheme is designed to merge the two orthogonal direction edge mappings, which has single response.On the work of edge detection, combined with wavelet transform, this thesis also proposes an X-ray image teeth positioning algorithm. The algorithm performs wavelet transform on the horizontal projection of X-ray image to determine upper and lower boundaries of the teeth area; on this basis, we get the edge mapping of a the specific region between the upper and lower boundaries, and then a method to calculate the threshold value is designed to process the vertical direction projection on the edge map image to get the left and right boundaries of teeth area. Finally we get the teeth area in the X-ray image. A large number of simulation experiments confirm the accuracy and effectiveness of the proposed algorithm. |
PDF Full Text Request