Image denoising with Grouplet transform

Grouplet transform (GT) can take advantage of the image´s geometry structure since the bases of Grouplet can adapt the different geometry structure in different scales. The association fields that calculate by The Block Matching algorithm which cannot adaptive to different textures cannot follow the turbulent texture contained in an image. Grouplet transform based on Streamline (GTS) introduced streamline to improve the performance of represent of turbulent texture. The starting pixel selected for association fields pruning was arbitrary, and one flow will prune to several flows that would destroy the original texture decreased the performance of Grouplet Transform. This paper proposed an advanced grouplet transform (AGT) that make use of the advantage of Greedy algorithm and Dynamic Programming algorithm in association fields pruning to ensure association fields well suited of the image´s texture structure. Experimental results show that the performance of image denoising by AGT-threshold outperforms GT-threshold denoising method and GTS-threshold denoising method.