Generalized multi-scale image decomposition for new tone manipulation

It is well known that the image multi-scale decomposition plays a foundational role in image processing and a variety of image multi-scale decomposition methods have been derived in various image processing tasks up till now. In many cases, the image decomposition needs to satisfy the conditions: edge-preserving or non-scale-hybridizing. Unfortunately, there has been no such multi-scale image decomposition approach satisfying both the two conditions simultaneously to our best knowledge. In this paper, one novel multi-scale image decomposition approach satisfying both the two conditions is proposed, which achieves different edge-preserving components under different scales. The proposed approach makes full use of well-designed adaptive mean envelope pixels to construct the segmentation envelopes to separate the different scales. Moreover, based on the multi-scale decomposition, an interesting multi-scale tone manipulation scheme and an effective new filtering algorithm for impulse noise are induced. Furthermore, numerical experiments compared with the existent methods are given to demonstrate the state-of-the-art performance of the proposed methods.