t—Edge detection in SAR images is a difficult task due to the strong multiplicative noise inherent to this imaging modality. Many researches have been dedicated to edge detection in SAR images but very few try to address the most challenging situations, namely edge detection in 1-look real SAR images. Motivated by the success of Convolutional Neural Networks (CNNs) for the analysis of natural images, we study the applicability of a classical CNN edge detector to SAR images, especially for edge detection in 1-look real SAR images. One crucial factor that contributes to the success of CNNs is a training dataset with labeled ground truth. In order to avoid the tedious job of annotating a large amount of SAR images, we simulate a SAR dataset leveraging the optical dataset BSDS500 [1] to train CNN models. Under the hypothesis that both optical and SAR images can be divided into piecewise constant areas, the main gap between simulated SAR images and real SAR images is that the possibility of mean intensity values for homogeneous areas is different. Therefore, we propose to train the CNN model on the gradient magnitude fields of the SAR images. The motivation behind this choice is that, provided a suitable definition of the gradient for SAR images, the gradient magnitude fields of homogeneous areas follow the same distribution regardless of their mean intensity values and the gradient distribution of two homogeneous areas across boundaries depends only on the ratio of their mean intensity values. The proposed stategy yields a detector with an approximately constant false alarm rate (CFAR property). We propose to train a state-of-the-art edge detector based on deep convolutional networks, namely HED (Holistically-Nested Edge Detector) [2], [3], on a Ratiobased Gradient operator (GR) on SAR images. The proposed edge detector, GRHED, achieves excellent performances in all our simulations including several 1-look synthetic edge images with different edge contrasts, two hundred 1-look optical images with synthetic noise, which are simulated from BSDS500, one synthetic SAR image and two 1-look real SAR images. In all these situations, the proposed edge detector outperforms concurrent approaches. The source code of GRHED can be found https://github.com/ChenguangTelecom/GRHED