Research Of Image Edge Detection Based On Wavelet Transform And EMD

The image edge detection is one of the basic contents in the area of image processing. When understanding and analyzing the image, edge detection is always be the first procedure and an essential attribute in detecting characteristics of images during the process of recognition and segmentation. Wavelet transform have good"time-frequency"and multiscale analysis technology, which makes it restrict noises while detecting edges. So it provides an effective way for image edge detection. However, considering wavelet base's various choices, inadaptability, and limitation, it is weak in analyzing unstable signals. Therefore, how to get high quality edge is important in image processing.This thesis mainly discusses about the applications of wavelet transform and Empirical Mode Decomposition (EMD) in image edge detection, the main work is as follows: The results compared and analyzed between using traditional edge detection algorithm and wavelet transform on edge detection; Image edge detection based on the combination of wavelet transform and two-dimensional empirical mode decomposition; Image edge detection based on the limited neighborhood two-dimensional empirical mode decomposition.Aimed at the problem that wavelet has weak ability to analyze non-stationary signal which has a bigger difference. Image edge detection based on the combination of wavelet transform and two-dimensional empirical mode decomposition was proposed. EMD is powerful to deal with non-linear and non-stationary signals. Driven by data and with the likeliness into 2D space, this thesis puts forward some improvements by using relevant algorithms, like, through dealing with leakage point issue by symmetry method, calculating sift terminate conditions by using the gray average of image data. Besides, extract image edge combined with wavelet. Simulation experiment has verified the validity of this algorithm.Aimed at the problem that the process of empirical mode decomposition consumes large amounts of time, an algorithm based on limited neighborhood empirical mode decomposition was proposed. Having abandoned the process of interpolation fitting surface and without limitation in terminal conditions, this method has greatly reduced the running time. Then applying this method to image noise reduction and image enhancement, afterwards, extracting the image edges. The simulation results verify the effectiveness of the proposed algorithm.