The Group Image Coding Algorithm Based On Multi-reference Image

In recent years,with the rapid development of science and technology,images and videos get a explosive growth and bring huge storage and management costs.Although the number of images is huge,there is a great correlation among images,especially in image sets with high similarity.The traditional image coding technologies like JPEG and MPEG can not make full use of these correlations and become powerless in improving the coding efficiency.Therefore,in order to further explore the relevance of the images,joint compression coding oriented the group image is studied which is called the group image coding.The traditional image coding technologies are introduced in the first place in this paper.Then,the group image coding is introduced,and pave the way for the group image coding based on multi-reference image.The key technologies of group image coding include generating coding structure,determining reference image and generating transformed images,and HEVC(High Efficiency Video Coding)baesed on BMC(block motion compensation).The theories and technologies of transformed images generation are described in more detail including deformation models generation,energy optimization,perspective transformations and photometric transformations.Although the discovery degree of the correlation among images of the group image coding using single reference image is larger than that of the traditional image coding,the correlation of multiple images in the image set is still not fully utilized.And now the researches are all focused on the group image coding based on single reference image.Therefore,the group image coding algorithm based on multi-reference image is put forward in this paper.There is an important problem in the group image coding algorithm based on multi-reference image which is determining multiple reference images of each image and generating transformed images of multi-reference image.The problem is related to the second key technology in the group image coding framework.This paper carries on the thorough research to the problem and proposes two new algorithms.Two new algorithms are named multi-reference image determination algorithm based on the determined coding sequence images and multi-reference image determination algorithm based on the determined coding sequence images and the current layer images.The reference images are searched only in the determined coding sequence images in the first algorithm.The second algorithm searches the reference images in the determined coding sequence images and the current layer images.The transformed images generation of multi-reference image is solved by the theories and technologies of single reference image in the group image coding.There is a brief introduction to HEVC at last.By testing some image sets,experimental results show that the group image coding based on the multi-reference image is separately 0.21 d B~0.49 d B higher than the group image coding based on the single reference image on PSNR(peak signal to noise ratio)under the same bpp(bits per pixel)when the number of reference images is 4 and the number of transformed images is 3 in these image sets.And the more reference images and transformed images,the higher the coding efficiency within a certain range.The experimental results show that the algorithm of multi-reference image determination in the group image coding can improve the coding efficiency effectively,especially for an image is associated with multiple images in the image set.