Pseudo-video Sequence Coding Structure Generation Algorithm For The Group Image Coding

As the image explosive grows in the cloud and the image resolution is getting higher and higher,how to compress the group image has become a serious problem.Traditional image compression using JPEG,JPEG2000 and other methods.Its main idea is that applying the correlation between pixels,the correlation between colours and the visual psychology of single image to image compression.However,there is usually some correlation between the group image,that is,there are similar scenes with different angles or same angles between images.For the group image,the correlation between the images is applied to codec processing,the efficiency of codecs will be futher improved and the storage space can be saved.The joint compression coding oriented the group image uses the correlation of images to generate pseudo-video sequence coding structure and encode it by video coding technology,which effectively improves the coding efficiency.In this paper,the framework of joint compression coding oriented the group image is introduced firstly,including the generation of pseudo-video sequence coding structure,the determination of reference image and the generation of transformed image,HEVC coding.The generation of the pseudo-video sequence coding structure relates to three modules: the distance calculation method between images,the image cluster algorithm and the generation of group of image coding structure.Then,this paper introduces the related technology of the group image coding in details.Among them,the distance calculation method between images introduces the existing method based on matching similarity.And on this basis,the matching area is introduced to describe the correlation between two images.In the image cluster algorithm,K-means clustering and AP clustering are introduced,and are compared in the later experiments.The generation of group of image coding structure is devoted to the map and the minimal spanning tree.Finally,the existing pseudo-video sequence coding structure generation methods are introduced,and the defects are analyzed.After clustering,The standard deviation of the distance from the cluster center to the other images reflects the difference in the internal correlation of the group of image.Aiming at the group of image that the internal correlativity has big difference,a depth-limited minimum spanning tree algorithm based on the shortest path is proposed.The idea of solving the shortest path to get the optimal reference path of each image is used.Aiming at the group of image that the internal correlativity has small difference,two new algorithms are proposed.One is the depth-limited minimum spanning tree algorithm based on the method of Zhu Liu and OTTC.The new edge is added to the minimum spanning tree obtained by the method of Zhu Liu and the depth-limited minimum spanning tree is futher acquired by OTTC.The other is the depth-limited minimum spanning tree algorithm based on the method of Zhu Liu and ant colony algorithm.After adding a new edge,the ant selects the starting point from the current nodes set and searches for the new node into the tree,then keeps updating the current nodes set until you get the minimum spanning tree that satisfies the depth limit.By testing the multiple image sets,the experimental results show that,if the internal correlativity of the group of image has big difference after image clustering,the depth-limited minimum spanning tree algorithm based on the shortest path has smaller weights and higher coding efficiency than the existing HAP algorithm under the constraint of each layer.In the group of image with small internal correlativitry,the depth-limited minimum spanning tree algorithm based on the method of Zhu Liu and OTTC and the algorithm based on Zhu Liu method and ant colony algorithm under the limitations of 3,4,and 5 layer have smaller weights and higher coding efficiency than the existing HAP algorithm.