Skip to main content

Multilateral Filtering for Image Denoising

Image denoising refers to removal of unwanted variations of pixels of the images. In this project, we present a novel iterative nonlinear filtering framework, termed multilateral filtering, based on the idea of generic local sim- ilarity. A set of local features is computed for each pixel using its local neighborhood. Two pixels are considered to be similar if the Euclidean distance between their corresponding feature vectors is small and vice versa. Multilateral filtering results in image smoothing while preserving edge and textural features. We extend the proposed framework to a multiresolution setting using both wavelet transform and Laplacian pyramid. Our experimental results show that the proposed methods produce comparable and often better results than the state-of-the-art denoising methods.

Experimental Results

Original Noisy Binarization result after Binarization result after
Image Image denoising with NLM denoising with MRM-Lap
  (14.20 dB)

(28.95 dB)

(33.65 dB)

Results of binarization after denoising for the Shapes Image

Related Publications

  • Irfan Talat Butt, Nasir M. Rajpoot, A Multiresolution Framework for Local Similarity based Image Denoising, Pattern Recognition (in press), 2012 [DOI]
  • Irfan Talat Butt, Nasir M. Rajpoot, Multilateral Filtering: A Novel Framework for Generic Similarity-based Image Denoising, in Proceedings International Conference on Image Processing (ICIP'09), Egypt, November 2009


    Test Images

    (Lena, Barbara, Monarch, Finger, Cosine Grating, Zebra, House)

    Project Code (Matlab & C++)