Efficient smart algorithms and architectures for real-time video transmission in pixel and frequency domains

Motion Estimation (ME) plays an important role in modern video coders since it consumes approximately 60%-80% of the entire encoder's computations. FSBB-ME is the main ME algorithm that is adopted by most coding standards. Although it provides the best coding efficiency amongst various ME algorithms, its computational complexity is very high. This work proposes several alternatives to reduce this bottleneck in video coding either in the pixel domain or in the frequency domain.;In Pixel Domain Video Coding, two systems are proposed. The first system is the AEST-APP-ME which is less complex and has less coding efficiency. AEST-APP-ME system achieves up to 99.39% saving in METS with degradation in PSNR by 0.12 dB and an increase in the bit-rate by 1.39% compared to the FSBB-ME. The second system is the DESST-DPP-ME-DISS-DESS which improves the coding efficiency with an increase in the computations. DESST-APP-ME-DISS-DESS system achieves up to 98.83% saving in METS with degradation in PSNR by 0.01 dB and an increase in the bit-rate by 0.69% compared to the FSBB-ME.;In Frequency Domain Video Coding, Further expediting of the ME process computations is accomplished by applying the previous proposed fast ME systems to the DCT-PC-ME. Simulation results reveal that there is an increase in PSNR by 0.48 and 0.71 and METS of 99.26% and 98.76% with reduction of bit-rate by 0.08 and 0.15 in case of using AEST-APPME and DESST-APP-ME-DISS-DESS systems, respectively, and compared to the FSBB-ME.;The main two components of the frequency domain encoder are implemented. The first component is the RT-DCT generator which generates all possible transforms required for the motion estimation in the frequency domain. Our architecture can easily perform real-time transformation for 4CIF video with 60 fps at 118 MHz. The second component is the MUE which is required in case of using dynamic padding search area. The proposed MUE can easily perform real-time manipulation for 4CIF video with 60 fps at 172 MHz.