Multiscale Product Edge Detection in Different Colour Spaces

Edge detection is an important task in image processing and becomes more complex for the colour image which provides more information in comparison to the gray image. In this paper it will be presented a colour edge detection approach based on the product of multiscale wavelet transform. The use of the product of multiscale transform improves the localization of detected edges comparing to the one scale wavelet transform. This method has been applied on colour images in RGB, YCbCr, HSV and YIQ spaces. The Berkeley University database BSDS500 was used to test this method. The comparison of the results has been done using ROC analysis parameters and the choice of the colour coordinates for extraction of significant edges in a colour image is discussed.