Algorithm For Image Insertion And Image Collection In Group Images

In recent years,with the development of science and technology,images appear explosive growth,and image quality is also increasing,this causes great pressure to the storage space and transmission network.Traditional image coding such as JPEG,JPEG2000,is mainly through the use of the redundancy between the image pixels or colors and the visual psychology to compress image.But there is usually some correlation in the huge number of images.Therefore,to take use of the correlation between images for higher image compression ratio,the group image coding appears.Because the group image coding compresses entire image set at a time,it is less flexible for adding images to image set or deleting images from image set than traditional image coding.It needs management of image sets to ensure the experience of users and saving storage space.This thesis first introduces the background and significance,then introduces the group image coding framework and the group image coding key technologies that includes generating coding structure,calculating transformation model,producing transform image and the key technology of HEVC(High Efficiency Video Coding).It requires to calculate the distance between the images、divide the images into different groups and generate the minimum spanning tree of image set for generating coding structure.Next,this thesis introduces the implementation process of the method that selects the optimum image set for inserted images basing on bag of words,including extracting the images' SIFT feature,establishing visual dictionary through k-means clustering,generating visual histogram of the image sets by visual dictionary,calculating the visual histogram of the inserted image and measuring the similarity of inserted images and image sets through visual histogram.Finally,this thesis analyzes the advantages and disadvantages of the existing coding structure adjustment method for image set which is inserted images into,and introduces a new coding structure adjustment method for image set which is inserted images into and a coding structure adjustment method which is used to merge image sets.The coding structure adjustment method for image set which is inserted images into will search the tree's branch of original image set layer by layer,to select the most similar reference image for every inserted image when satisfying the depth limitation.The coding structure adjustment method which is used to merge image sets,will use the relationship between the root node of the non-main image set and the images of main image set to merge.By testing some image sets,the experimental results show that the methods that selects the optimum image set for inserted images based on bag of words,can choose the most similar image set which is same with the subjective judgment for inserted images;The cost of the coding structure generated by the new coding structure adjustment method for image insertion proposed in this thesis is less than the existing coding structure adjustment method based on root node.Using the coding structure generated by the new coding structure adjustment method for image set which is inserted images into to encode image sets can get higher coding efficiency than the existing coding structure adjustment method based on the root node;The coding structure adjustment method which is used to merge image sets can merge the image sets with low complexity,the merged image set is not only more convenient for retrieval,but also have better coding efficiency than the image sets which are not merge.