Design & Development of Bit-Stream Containing Multi-view Video & Associate Depth Maps - Enabling 3-D Video

Abstract

In this thesis, a new method is proposed to EncodenDecode multi-view texture and associate depth maps in to a single bit stream. Modi ed encoder sends modi ed bit stream to the modi ed decoder which is smart enough to sense the presence of depth maps in to a bit stream and able to decode texture and depth maps seamlessly. Encoded depth maps are merge in to an associated texture using conventional steganography approach with a very less loss in PSNR of texture, 0.1 dB to 0.2dB for high delay and 0.2 dB to 0.5 dB for low delay.The proposed change in the existing algorithm is able to handle the bit stream with and without depth information embedded in it.Modi ed encoder uses unused bit in bit-stream to signal the modi ed decoder about presence of depth map, so decoder can easily sense presence of depth in a bit- stream. Encoder accommodates the depth map bits in video coding layer using steganography and remaining bits in the Network abstraction layer. Further proposed the method, in which decoder will extract the entire merged depth bits from the modi ed bit stream. After accommodating all the depth maps in to associated texture, modi ed bit- stream send to the the modi ed decoder. It will decode depth and texture from the bit-stream.It decodes depth maps according to Multi View Coding (MVC) standard.Modi ed decoder will run in a multi pass fashion.In rst pass it will decode texture and extract depth maps while in second pass it will decode depth maps. It is backward compatible to MVC standard bit streams as well as our modi ed stream. Main aim of this work is to generate single bit-stream so that minimum changes are required in conventional decoder.It requires to add pre processing block on encoder side while post processing blocks on the decoder side.Core design is same as the H.264nMVC encoder/decoder; Designer needs to add some functional block for accommodating new changes. Benchmarking results with HD sequences has been reported in this thesis. This technique is very useful in a Free view point video (FVV) or 3-D video (3DV).