The improvement of a feature-based image mosaics algorithm
Xiaoru Zhang · 2008
In image mosaics,it is very hard to improve both the quality of mosaics and the runtime. This paper discusses the two key parts of image mosaics to solve these problems:feature matching and image blending.In feature matching,a traditional method to purify feature points is RANSAC algorithm in feature-based image mosaicing.To prevent its low efficiency,this paper presents a pre-pure feature points method by clustering to enhance it.In image blending,to de-ghost and present a natural stitched picture,optimal route and brightness weighting function methods are ac- tualized.Experiment results show that after improving the two parts of image mosaics algorithm, the efficiency is greatly enhanced and the ghosting is removed effectively to make the blended pic- tures inartificial.