Research On The Mural Image Mosaic Algorithm Based On Local Invariant Feature Extraction

Mural,also known as the art on the wall,is the treasure of ancient Chinese culture.It is one of the earliest form of painting in history of mankind.Murals contain lots of information about history,humanities and arts,which are valuable.Getting high resolution panoramic mural images plays an important role in the process of murals’ protection.So the investigation of mural images mosaic algorithm has great significant for the protection of murals and other ancient relics.Image matching,which directly decides the result of image mosaic,is one of the key steps in the image mosaic.Among many image matching algorithms,SIFT algorithm is widely used due to its robustness.But its performance is poor.Since ancient mural images have rich local features and cracks,and Harris algorithm is efficient in calculation,we propose a mural image mosaic algorithm based on local invariant feature extraction in this dissertation,which includes the following two aspects:(1)A mural image matching algorithm based on Harris-SIFT has been proposed.In order to improve the efficiency of the SIFT algorithm,the proposed algorithm uses the Harris corner extraction algorithm in local extreme points extraction in murals images,to generates the structure-optimized and dimension-reduced SIFT descriptor,and finally realize the image matching between mural images.The experimental results show that,compared with the traditional SIFT algorithm,the proposed algorithm is able to ensure the stability of image matching and improve the efficiency of the algorithm.(2)A mural image mosaic algorithm based on local invariant feature extraction is proposed in this dissertation.Since mural images have rich local features with some cracks,the idea to use local image to represent the whole is introduced in the algorithm.With the semi-automatic crack extraction technology,the proposed algorithm takes the extracted local image which contains some cracks as the local feature of the entire image,then matches the entire image by the local images,and gets the transformation matrix between local images.The algorithm takes the local image transformation matrix as the full image transformation matrix,and finally realizes the mural image mosaic.The experimental results show that the algorithm is feasible and robust.