One easy, but effective approach uses a minimum pixel value filter for the
downsample process which ensures a big compression rate, when used with Huffman coding for compression.
For further verifying the effect of feature dimension to recognition accuracy, we
downsample the face images to 10 x 10, 20 x 25, 25 x 30, 35 x 40 and 40 x 50, respectively, which correspond to feature dimension of 100, 500, 750, 1400 and 2000.
The recognition rates when the ear is occluded using Gabor orientation + NSRC,
Downsample + SRC, Random ear + SRC, and Eigenear + SRC are illustrated in Figure 9.
According to the RSSI measuring frequency, we
downsample the estimated real VO routing path.
The second step is to
downsample the right CCD image to be matched and target area template image extracted from the left CCD image.