Research And Improve On Canny Algorithm For China Paintings

Image edge detection is not only the precondition of image analysis, but also a very important step, The effect in edge detection directly affect the follow-up studies. In this paper, the method of painting edge detection is analyzed and studied, which providing the premise for the subsequent three-dimensional modeling or digital simulation based on paintings. The main reasons for many noises and low-quality image of ancient paintings are the results of long-time storage and careless,we can hardly obtain satisfied result by the traditional edge detection methods. So this algorithm,in this paper,based on the classic Canny algorithm be improved, main works is as follows:1. The traditional edge detection operator and the classical Canny algorithm advantages and disadvantages are analyzed, and the experimental results showed that the traditional edge detection algorithm and the classical Canny algorithm are not applicable to edge detection of paintings.2. The preprocessing of ancient paintings for its many noise and quality is not clear. It is necessary to reduce the influence of noise before the edge detection. The guide filtering method to process the noise of the image of ancient paintings, which instead of Gaussian filter of the classical Canny algorithm.3. Improving the classical Canny algorithm. First, for the reason of that the calculation of gradient amplitude of the classical Canny algorithm is more sensitive to noise. Second, for the reason of that the number of reference pixels in the process of Non-Maximum Suppression of the classical Canny algorithm is too few. Through the improvement of these two points, the edge positioning is more accurate, and the effect of noise on edge detection is reduced.4. In view of the difficulty of the classical Canny algorithm in threshold selection, then the steps of the two improved adaptive threshold selection algorithms are described in detail. Finally,through the comparison of the experimental results, the effectiveness of the improved method in this paper and two adaptive threshold algorithm is verified in the paintings.