CAJ | 학술논문

给出了一种特别适合于低功耗硬件实现的运动和图像特征自适应的块匹配运动估值算法，它采用一种新的由运动跟踪搜索和望远镜搜索组成的两阶段可提前中断的搜索方法，根据宏块的运动特征和图像特征自适应地控制中断、调整搜索窗的大小和改变像素数据的表示精度。模拟结果表明新算法的平均运算量仅为传统算法的30％～40％左右，却可以得到与传统算法相同的视频解码图质量。一种一维心动阵列保证了这一新算法的低功耗硬件实现。
급출료일충특별괄합우저공모경건실현적운동화도상특정자괄응적괴필배운동고치산법，타채용일충신적유운동근종수색화망원경수색조성적량계단가제전중단적수색방법，근거굉괴적운동특정화도상특정자괄응지공제중단、조정수색창적대소화개변상소수거적표시정도。모의결과표명신산법적평균운산량부위전통산법적30％～40％좌우，각가이득도여전통산법상동적시빈해마도질량。일충일유심동진렬보증료저일신산법적저공모경건실현。
A new adaptive block-matching algorithm is proposed for low power im plementation of motion estimators. The algorithm fully utilizes the motion and image characteristics to dynamically reduce the computational complexity. A novel two-stage search strategy is used, in which a motion track search is conducted followed by a telescopic search. The two-stage search can end early after the m otion track search and the search window size is adaptively adjusted. In additi on, the bit-resolution of the pixel values adaptively varies with the image cha r acteristics and alternatively varies with the telescopic search steps. Simulatio n results show that, compared with the conventional algorithm, the new algorithm reduces the average computational load to only 30%～40% while providing similar decoded picture quality. A systolic array search engine is proposed for low-pow er hardware implementation of the algorithm.