An edge detection based noise removal algorithm

• We introduce an edge detection based noise removal algorithm.
• The denoised image is a linear combination of the original image and its averaged value.
• The coefficients are computed by a gradient-based edge detector in an exponential form.
• The control parameters are adaptively computed with noise variance.
• The method is easy, fast, comparable with ROF filter and useful for very noisy images.

We present an image denoising method using the edge map of an image. The denoised image is considered as a linear combination of the observed image and its average value, where the coefficients are controlled by a local edge detector. The parameters are set on suitable values related to noise energy computed by the curvature norm of the original image. Implementation can be done in a single iteration and the speed of the process is reasonably high. Noise reduction quality of the introduced method is compared with Wiener and Total Variation based filters for some images. The method appears to be easy, fast and useful for very noisy images. The differences between our method and the patent 6229578 “Edge Detection Based Noise Removal Algorithm” are explained.

Figure optionsDownload as PowerPoint slide