| 摘要 |
A system and method for recognizing image sequences. A video decoder having a raster/cluster converter and an image carousel having a memory organized as a shift register are connected to each other by a discrete cosine transform (DCT) transformer. The system extracts characteristics from individual images of an image sequence provided by the video decoder based on the brightness of the images. A correlator connects a reference memory to the image carousel such that the image sequence from the video decoder is supplied to the correlator as a characteristic vector and combined with a reference sample for storing in the reference memory. The system decorrelates the characteristics by a quasi-stochastic characteristics extraction over a plurality of the images, digitizes the characteristics, and compares the digitized characteristics to the reference sample.
|
