Algorithms On Edge Detection Based On Gradient Operators And Otsu

The detection of image edge is the basis of image analysis and understanding, and it is the important premise of the image feature extraction and target recognition. The target image understanding and perception of the computer is influenced severely by the integrity of the results. Therefore, the detection of image edge plays a crucial role in computer vision and digital image processing fields.On the basis of the first and second order differential operator and Canny operator, we focused on analyzing the edge extraction of the underwater particles. According to the underwater imaging characteristics of underwater suspended particles, we proposed the contributions as follows.(1) As the Sobel operator has some problems of extracting the underwater particle edge in complex background, such as rough edges, low threshold accuracy, we presented an algorithm of image edge detection based on both threshold selection and gradient operator. We employ the Newton iteration to quickly converge to the optimal threshold. The experimental results show that it has high accuracy in detecting underwater particles'edge, and its contour is clear and accurate.(2) As underwater illumination results in the noise, such as horizontal stripes, which is not robust to the traditional gradient operator due to high sensitivity to detecting them, we proposed a new algorithm based on both two-dimensional Otsu (2D-Otsu) and image edge enhancement. It adopts the gradient sharpening operator to eliminate the noises of horizontal stripes. The experimental results show that it can detect the edge of the underwater particles and filter the noise, such as horizontal stripes, and it is more robust to horizontal stripes.(3) As traditional Canny operator is manually determined in the threshold and less adaptive, we presented an algorithm based on both two-dimensional Otsu and Adaptive Threshold in Canny for edge detection. The experimental results show that it can solve the contradiction between the noise filtering and edge maintenance. Therefore, it has strong robustness and adaptability.The new algorithms in edge detection can solve some problems in extracting the edge of underwater particles for conventional gradient operator. They have more advantages in detecting precision, precise positioning, adaptability and robustness.