A Pixel Orientation and Adaptive Search Range based Complexity Reduction in H.264 Scalable Video Coding (original) (raw)

H.264 advanced video coding (AVC) is prolonged to scalable video coding (SVC) for all categories of network transmission. Due to the adoption of exhaustive full search method, SVC computational complexity (CC) has increased when compared to AVC. To reduce the computational complexity, various fast mode decision (FMD) algorithms have been developed. In FMD algorithms, the selection of mode in macro block (MB) needs an efficient algorithm to reduce computational complexity. These FMD algorithms reduce Computational Complexity in saving time alone with no increment in peak signal to noise ratio (PSNR) or no detriments in bit rate. In other words, the existing FMD algorithms are effective in one measure viz., time saving as a reduction of computational complexity but doesn't affect PSNR and bit rate. This research paper describes how a pixel orientation algorithm (POA) executed for SVC will result in reduction of computations complexity in all three measures viz., time, PSNR and bit rate compared with the all existing algorithms. From the results, the overall performance of POA resulted in 3.76 % faster time, 1.43 dB increments in PSNR and 0.45 kbps detriments in bit rate as compared to joint scalable video model (JSVM). Keywords-SVC, Computational Complexity, pixel orientation mode, PSNR, Encoding Time I. INTODUCTION A video sequence is quieter of a delay-sequenced order of still pictures. The uncompressed video is of abundant size and for the reason of storage and transmission, compression is much needed. Scalable video coding (SVC) has been decided as an extension of the H.264/ advanced video coding (AVC) standard. Distinguished to the former video coding standards, SVC is set to encode the signal once, but the bit stream can be partly removed relying on the particular rate and resolution, and the resultant sub-stream will be decoded by any decoder [1]. The sub-stream can be scaled over and over until the smallest description; also called base layer (BL). BL is the lower layer having low resolution, low frame rate and low quality and backward compatible with H.264/AVC. The lower layer can be enhanced by optionally adding the additional bit streams from one or more higher layer are referred as enhancement layer (EL). SVC backs three greater scalability, especially temporal, spatial and quality scalabilities. For temporal scalability, the subset of the bit stream represents similar content however with a different frame rate. Temporal scalability in SVC is achieved by employing a structure of hierarchical B picture prediction (HBP) [1]. A video is a collection of I, P and B frames referred as Intra frame, Predictive frame and Bi directional Predictive frame. I frame also called as an Instantaneous Decoding Refresh (IDR) which uses intra prediction due to its more importance information producing low coding efficiency. P frame are coded with inter prediction which uses previously coded I or P frames of temporal BL. I or P frames of BL are referred as key frames which is the first frame of a Group of Picture (GOP). A GOP is a group of frames between two key frames and may be in the order of 4, 8, 16, 32, etc., B frames are coded using I or P frames which is of less importance information producing high coding efficiency. For spatial scalability, the subset of the bit stream denotes similar content, however with a different frame size. It includes new tools like inter layer prediction for better coding efficiency. Various inter layer prediction such as intra, motion and residual used to reduce redundancy [1]. Spatial scalability in SVC is performed by a multi-layer approach. Every layer has a distinct size and referred as dependency identifier D from spatial base layer to spatial enhancement layers [2]. In every spatial layer, the encoding approach for the single layer coding like the motion-compensated prediction and intra prediction are involved to increase the coding efficiency. Further two inter-layer prediction concepts [3] are added in SVC: MB mode prediction and associated motion parameters and residual signal prediction. For quality scalability, the subset of the bit stream will offer a similar spatio-temporal resolution, however with variable quality. Scalability in combination with the above one or more forms can be referred as combined scalability or hybrid scalability. However, with its inherent characteristics SVC is scalable and efficient in coding compared to the previous video coding standards at the cost of increased complexity in computation. The computation complexity (CC) of SVC limits its involvement in global applications. If SVC adopted globally then there is a need for a significant decrease in CC. The CC includes coding of difference information, desired mode and motion. The mode predicted in SVC based on intra and inter layer coding. Intra coding involves the difference information between the current macro block (MB) and previous MB, whereas inter layer coding takes the information from the previously coded MB of lower layer (BL) in addition to intra coding. Conventional JSVM full search mode will calculate rate distortion cost (RDC) for all possible modes and chose the best mode with minimal RDC. But this method consumes IEEE ISBN No. 978-1-5090-4558-7 802