The Study Of Uncertainty About Image Edge

The edge of an image carries the majority information, and therefore, image edge detection takes an extremely important position in digital image processing. Both the actual image and its edge have the uncertainty. A quantitative guidline which is for the evaluation of the effectiveness of edge detection is proposed in this paper, which is from the point of view of uncertainty. And take the image edge uncertainty as the target, the parameters of the Canny algorithm are optimized. Analyzing the uncertainty of the edge detection algorithm has great significance to evaluate the accuracy of the edge. The main research contents in this paper are as following:1. The paper first introduces the uncertainty of an image, calculates the uncertainty of the image with local standard deviation method, and improving this method. The effectiveness of the improvement is verification by experiments. In order to assess the performance and quality of the edge detection algorithm, the conception of image edge uncertainty is defined. And two kinds of methods which calculate the image edge uncertainty are also defined according to the classification of the image.2. Monte Carlo method is used, adding noises to the ideal image, simulated reality image, and the uncertainty of the edge is quantificationally analyzed. Extracted the edge using Canny edge operator, to calculate the uncertainty of the image edge. And the uncertainties of the edge under four different kinds of algorithms are compared by experiments. Take uncertainty as a guidline to evaluate the precision of the algorithms. The method of calculating the edge uncertainty is feasible, which is proved throgh comparative experiments.3. The filter parameters and the threshold values of the Canny detector are optimized in this paper. Under the different noise conditions, take uncertainty as a guidline, the optimal filter parameters and thresholds are searched for.The study results in this paper show that using uncertainty to evaluate the image qualiry and the effectiveness of the edge detection is feasible and effective. The conclusions which are obtained through the size of the uncertainty are identical with that of the subjective evaluation. On this basis, take uncertainty as a guidline, the parameters of the Canny operator are optimized, as a result, the optimal parameters under different noise conditions are obtained, a preliminary exploration is made in this paper to optimize operator's parameters.