Abstract We present an efficient and improved method for hair removal from dermoscopic images, which is faster and can remove hairs more effectively as compared to the existing and widely used DullRazor ®. To do so, we first detect the predominant orientation of hairs in the image by using Radon transform, followed by filtering the image by Prewitt filters using the orientation of existing hairs. Undesirable effects, such as non-hair structures and noise are removed from the image by thresholding–averaging–thresholding, followed by smoothing. The proposed scheme has the advantage of removing bubbles, as well. Implementation of our proposed scheme on different dermoscopic images validates its improved performance.