Implementation And Evaluation Of Residual Color Transform For 4:4:4 RGB Lossless Coding
Agawane, Pooja Vasant
MetadataShow full item record
The 4:4:4 video sampling format promises an excellent quality video. It is gaining a lot of attention due to its significance in the professional applications of multimedia processing. The use of the RGB color space for video processing is attracting, both, the industry and the academia. Extensive research is being undertaken to achieve better compression efficiency and high coding gain in the RGB (red, green, blue) color space. In contrast to typical consumer applications, high quality video is demanded in areas such as professional digital video recording, video post production and digital cinema. These latter applications demand all three color components to be represented with identical spatial resolution. Some of these applications demand that each color component of the video signal be captured and displayed with a precision of more than 8 bits per sample. Most of the time, a video signal is captured and displayed in the RGB color space. During the transition phase between the capture and the display of video, encoding and transmission takes place. RGB color space is not an optimum choice for coding and achieving compression. This is because of the significant amount of statistical dependencies between the red, green and blue components of the given video signal. In order to take advantage of these statistical properties, a decorrelating transformation from the RGB color space to some other suitable color space is applied. The various standardization bodies, for example, ITU or SMPTE have defined several color transforms for video coding purposes. One of the color spaces is denoted by YCbCr. This color space includes one luminance component (Y) and two chrominance or color difference components (Cb and Cr). The captured video signal is transformed from the RGB space to YCbCr space. However, this conversion has its limitations. This conversion includes the use of decimal coefficients. Since the samples of a video signal are represented using integers, rounding errors are introduced. Also, in order to achieve high coding efficiency, the complexity of the transform is also increased. To overcome these limitations, the Fidelity Range Extensions (FRExts) amendment of H.264 supports a new color space. This is the YCgCo color space where Y stands for luminance, Cg stands for green chroma and Co stands for orange chroma. The principle of the residual color transform is also introduced in the FRExts. It exploits the redundancy among the residual data of each RGB component after intra/inter prediction. As a part of this research, the RGB to YCgCo color space transform is applied to the residual data. This thesis aims at the implementation and the evaluation of the residual color transform. This transform is applied to high definition sequences, with the resolution of 1920x1080. The YCgCo color space has improved coding gain relative to both RGB and YCbCr.