The studies and application of quaternion convolution in the color image filter

• Main Authors: Lin Jia-Nan, 林佳楠
• Other Authors: Hsu Chau-Yun
• Format: Others
• Language: en_US
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/95119587767797345675

碩士 === 大同大學 === 通訊工程研究所 === 89 === This thesis presents two new filters for edge-detecting filter and sharpening filter, which are based on a new type of image convolution in which color images are convolved with a pair of quaternion-valued (hypercomplex) masks. When new edge-detecting filter applied to a color image, the new filter converts areas of smoothly varying color to shades of gray, and generates colors in regions where color edges occur in the image. The result may be easily thresholded for chromaticity to produce an edge image exhibiting color only where the original image exhibited a sharp change of color. A LoG-type filter based on quaternion convolution is developed in this thesis. It is isotropic, and the characteristic of rotation about the grayline of RGB space is applied to detect the edges of circuit on PCB. A new Sharpening filter applied to color image that contrast enhancement occurs in high detail color areas and little or no image sharpening in smoothly varying color areas. This sharpening filter is more robust to the present of noise in the input images than traditional sharpening the separate R, G and B channel approach.