Classification Based Adaptive Vector Filter for Color Image Restoration
Zhonghua Ma, Hong Ren Wu · 2006
A new adaptive vector filter is proposed for restoring color images corrupted by impulse noise. The local image structure is estimated by a series of central weighted vector median filtering operations. Then a classification process is applied to map the local estimate errors into a group of mutually exclusive structure partition cells. For each partition, an optimal weighted filter is applied to provide the best image structure restoration. The new filter has demonstrated satisfactory results in suppressing various distinct types of impulse noise. Noticeable performance gains have been demonstrated over other existing methods in terms of objective measures and perceptual quality.