Plenary lecture 1: impulse noise removal with polynomial interpolators
Cheng‐Hsiung Hsieh · International Conference on Signal Processing · 2010
This plenary speech presents an impulse noise removal approach which employs boundary discriminative noise detection with boundary resetting (BDNDBR) and polynomial interpolators. In the proposed approach, two stages are involved: noise detection and noise replacement. The noise detection performed by the BDNDBR is to identify a noisy pixel in an image. If a pixel is noise-free, then keep it intact. Or replace it with uncorrupted neighborhood pixels through the polynomial interpolators. Note that miss detection happens in the well-known BDND scheme when the noise density is high. The miss detection is even worse for cases with unbalanced noisy density where the portions for salt noise and pepper noise are different. To avoid the miss detection, a boundary resetting scheme is incorporated into the BDND. By this doing, the problem of miss detection in the BDND is prevented. In the noise replacement stage, two polynomial interpolators are adaptively selected to replace a noisy pixel according to the noise density. In the cases with higher noise density, a zero-order polynomial interpolator called adaptive nearest neighbor interpolator (ANNI) is used while a first-order polynomial interpolator called adaptive linear interpolator (ALI) is employed for the cases with lower noise density. Several examples are provided to justify the proposed BDNDBR, ANNI, and ALI. Moreover, the proposed noise removal approach is compared with other reported approaches as well.