The human visual system exhibits reduced sensitivity to distortions in the regions of an image where the rate of change is significant. This entails that a watermark with significant value can be robust and if it resides near around edges and textured areas of an image, it would be imperceptible as well. The present work exploits this characteristic of the Human Visual system to embed a robust and imperceptible watermark in transform domain using edge detection. The embedding is done at block level in either the Discrete Hartley Transform (DHT) domain or in Discrete Cosine Transform (DCT) domain. The watermark is embedded block by block in different blocks of the image. The decision whether to embed in DHT domain or in DCT domain is based on the number of edges which exist in a given block in the image to be watermarked. Hence the threshold number of edges acts as a key in this algorithm and is used in the embedding as well in the extraction process of the watermark. The results demonstrate the robustness of scheme against common image processing operations like cropping, low pass filter, noise and lossy JPEG compression with various quality index factor. Results also illustrate that the watermark is perceptually transparent, recoverable, recognizable and robust even after the watermarked image has been passed through severe attacks.