ENCODING A VIDEO SIGNAL WITH HIGH RESOLUTION ENCODING FOR REGIONS OF INTEREST

摘要

Method and arrangement for encoding a video signal, wherein a selection (<HIL><PDAT>39</BOLD><PDAT>) is made between the generation of a motion-compensated encoded video signal (I, B, P) and the generation of a higher-resolution picture (I<HIL><PDAT>H</SB><PDAT>). The creation of the higher-resolution picture (I<HIL><PDAT>H</SB><PDAT>) relies on estimating motion in a series of subsequent pictures (f<HIL><PDAT>1,2,3, </SB><PDAT>. . . ) of lower resolution. The higher-resolution picture (I<HIL><PDAT>H</SB><PDAT>) is interpolated (<HIL><PDAT>38</BOLD><PDAT>) from these pictures (f<HIL><PDAT>1,2,3, </SB><PDAT>. . . ). In a preferred embodiment of the invention, previously determined regions of interest (S(ROI)) are selected (<HIL><PDAT>36</BOLD><PDAT>) from the video signal. The pictures relating to these regions of interest (f<HIL><PDAT>1,2,3, </SB><PDAT>. . . ) occupy less memory (<HIL><PDAT>33</BOLD><PDAT>) than entire pictures as picked up by an image sensor (<HIL><PDAT>2</BOLD><PDAT>). A conventional picture memory (<HIL><PDAT>33</BOLD><PDAT>) is sufficient to store the series of pictures (f<HIL><PDAT>1,2,3, </SB><PDAT>. . . ) that relates to a certain region of interest. Motion estimation (<HIL><PDAT>31</BOLD><PDAT>) is performed on these pictures and their motion vectors (m) are stored in a vector memory (<HIL><PDAT>37</BOLD><PDAT>). The higher-resolution pictures (I<HIL><PDAT>H</SB><PDAT>) are interpolated (<HIL><PDAT>38</BOLD><PDAT>) from the stored pictures (f<HIL><PDAT>1,2,3, </SB><PDAT>. . . ) and their motion vectors (m). The invention is applicable for sensor (<HIL><PDAT>2</BOLD><PDAT>)-encoder (<HIL><PDAT>3</BOLD><PDAT>) integration, for example in a camera system (<HIL><PDAT>1</BOLD><PDAT>).</PTEXT>