THE IMPLEMENTATION OF REFERENCE FRAME PREDICTION FOR REDUCING MEMORY BANDWIDTH AND COMPUTATIONAL COMPLEXITY IN EMBEDDED SYSTEM

摘要

PURPOSE: A method for compressing a multiple reference image for reducing a memory band and the complexity of calculation in an embedded system is provided to reduce a use amount and a calculation amount of a memory without lowering of compression efficiency. CONSTITUTION: A converter(3) receives an initial frame of a video signal through a switch(SW) for converting the first frame into appropriate data. A quantizer(4) quantizes the data for outputting the same to a video multiple coder. An inverse quantizer(5) inversely quantizes a quantization index for a conversion coefficient. A restoring unit(6) restores the data for restoring the first frame of the video signal. An adder(7) outputs the first frame and adds the previous frame to the current frame. An additional picture memory stores output of the adder. A subtractor(1) subtracts a frame of a picture memory(9) previously stored from the current frame for extracting motion information. The converter converts the motion information. The adder stores the motion information to be used as a reference image. A background area memory(10) detects motion and preliminarily stores a background area revealed according to the predicted motion.