| Summary: | Edge detection plays an important role in image processing. Edge detectors have always been a compromising between information and noise. Since edge detection is a derivative operation, it tends to amplify noise. This means that increasing the amount of information may increase the noise as well. There are a variety of edge detectors or operators with different size of the kernel. In general, many established edge detectors focus on the gradient in grayscale image to detect edges. This paper proposed an improvement of edge detection algorithm by considering two edge features: gradient and length. In the proposed algorithm, the threshold value of the gradient was set to a value similar to a default value used in other existing edge detectors. The length of the edges feature was used to increase the robustness of the proposed algorithm towards the noise. The proposed algorithm was validated with synthetic and natural images with the inclusion of three types of noise: additive, multiplicative and impulsive noises. Results were compared with other established edge detectors whereby the proposed algorithm demonstrated its superiority in handling edges in low contrast regions and less sensitive towards the noise.
|
|---|
