Research On Image Edge Detection Based On Wavelet Transform And Its Applications

Edge detection is quite important in image processing, since edges usually include the main information of an image. Edges always exist in the anomalous constructures and unstable locations which show the positions of the silhouette of an image. The silhouette is usually the most important characteristic of an image, so it is necessary to detect edges.There have been many algorithms proposed for edge detection of images, but the existing theories and algorithms still have some drawbacks and could not detect edges efficiently in some cases. Therefore, it has become the principal aspect to design new methods for specific application requirements or to find advanced algorithms for existing ones to obtain satisfied results.Wavelet transform has superior characteristics in time-frequency domain and is also effective in analyzing the singularity of signals, so in this dissertation the applications of wavelet transform are studied.The traditional edge detection algorithms are sensitive to noise, and it will be difficult to detect the edges precisely on a signal scale. Combining multiscale analysis of wavelet transform, a new method is proposed in this dissertation based on Quadratic B-spline wavelet transform to overcome these two drawbacks. In this algorithm, above all, Quadratic B-spline wavelet is proved as the optimal operator quantitatively and the coefficients of Quadratic B-spline wavelet filters are obtained through algebraic methods, then an advanced algorithm is proposed based on adaptive thresholds by comparing with the traditional algorithms. It is showed that, in gray image edge detection, this algorithm is fast, precise and more efficient in getting subtle edges by analysis on multiscale.The color image edge detection is the core section of image edge detection. In this dissertation, the algorithm of gray image edge detection is introduced for color image edge detection and based on multiscale fundamental form of multivalued images, a new expression of color image gradients is proposed, which combines the different information among the three bands of the HSV color model. The experimental results show that, this method has excellent ability in deleting the redundant information of the three bands and intensifying the precision.