Title :
Three-Dimensional Point-Cloud Plus Patches: Towards Model-Based Image Coding in the Cloud
Author :
Dong Liu ; Tong Shao ; Houqiang Li ; Feng Wu
Author_Institution :
Key Lab. of Technol. in Geo-spatial Inf. Process. & Applic. Syst., Univ. of Sci. & Technol. of China, Hefei, China
Abstract :
In the era of cloud computing, there are many correlated images in the cloud, joint compression of these images may provide much higher compression ratio than individual coding. Model-based coding is an appealing approach to image coding in the cloud, as it removes knowledge redundancy among images that share the same model. In this paper, we make an attempt to model-based image coding for landmark images, where our model consists of three-dimensional (3-D) point-cloud plus image patches to describe the geometry and surface color of the landmark respectively. The camera parameters of an input image are estimated based on the 3-D point-cloud and the patches in the model, and then prediction image is generated by selecting, warping, and stitching image patches as well as illuminance compensation, the residue between original and prediction images is compressed by P-frame coding in HEVC encoder. We perform experiments on an Internet photo collection to verify the effectiveness of the proposed scheme. Preliminary results display the superior performance of our scheme that achieves as high as 39.9% bits saving compared to HEVC intra on a single image. The proposed scheme indicates a promising approach to image coding in the cloud and is worthy of in-depth investigation.
Keywords :
cloud computing; data compression; image coding; image colour analysis; image sensors; lighting; HEVC encoder; HEVC intra; Internet photo collection; P-frame coding; camera parameters; cloud computing; compression ratio; illuminance compensation; image patches; joint compression; knowledge redundancy; landmark images; model-based image coding; prediction image; surface color; three-dimensional point-cloud plus patches; Cameras; Decoding; Feature extraction; Image coding; Lighting; Redundancy; Solid modeling; 3-D Model; Cloud Storage; Image Coding; Model-Based Coding; Point-Cloud;
Conference_Titel :
Multimedia Big Data (BigMM), 2015 IEEE International Conference on
Conference_Location :
Beijing
Print_ISBN :
978-1-4799-8687-3
DOI :
10.1109/BigMM.2015.15