Edge detection in digital images using small mask and vector operators and the method of polar histograms
An illustrated comparison of ten small mask edge.detection systems is presented. Each operator was applied to three test images and a fixed percentage of pixels was assigned to the edge category through a system of global thresholding. A new nonlinear 3 x 3 operator was developed which is sensitive to edges which separate regions of constant gray levels from regions of irregular gray levels.
Secondly, systems for estimating the gray level gradient as a vector are described together with methods for evaluation of vector accuracy. A histogram is formed by calculating the sums of the vectors which fall into narrow angle bands. This histogram is drawn in polar coordinates after low pass Butterworth filtering and used to infer characteristics of the original images. The polar histogram is used to estimate the relative angle of orientation of two images which were rotated with respect to each other prior to digitization.
The polar histogram is also used to estimate the number, orientation, and importance of principal axes in images of stereological interest.
Thesis78D3942.pdf
25.02 MB
Adobe PDF
9163c15b2b09642e265e01f2f9e0c7d4