| Summary: | 碩士 === 國立宜蘭大學 === 電子工程學系碩士班 === 102 === In recent years information booming faster, and a lot of identifications gradually become the top research field, such as face recognition, license plate recognition, fingerprint recognition, etc. However, these identifications may suffer interference from noises, and the wrong detection will be occurred. Because the edge is one of the basic characteristics to identify the object shape and structure, and it is also the basis of image analysis and recognition. In order to make the image edge detection not suffer interference from noise, this thesis presents a stationary wavelet transform method for edge detection. First we use stationary wavelet transform to decompose Gaussian noise image, and it will generate horizontal, vertical and diagonal high frequency coefficients for three scales. And then Rosin thresholding is applied to make decisions for edges. Finally, we use PFOM measure to assess the proposed algorithms, sobel, canny, LoG and a’trous algorithm edge detection methods.
|
|---|
