In this paper, we propose a Scale-Invariant Salient Edge Detection framework (SISED) using Hadamard-Product operation. Current HED-like edge detection approaches fuse multiple side outputs to produce the final edge map, which contains noise and unwanted edge details. Scale-invariant salience provides strong and convincing evidence for precisely describing object contour. The normalized Hadamard-Product is able to find and extract scale-invariant features by multiplying a set of side outputs. The computed Scale-Invariant Salient Edge (SISE) maps capture the hierarchical structure of contour details and can be utilized to improve the detection accuracy. The experimental results show SISED reaches state-of-the-art performance.