Abstract: Method and apparatus for real time recognition of test images by comparison with sample images. The test image to be recognized is scanned to produce a plurality of analog signals of the luminance levels. These luminance signals are screened to eliminate background signals and are converted to digital signals. A distribution function of the occurrence frequency versus the luminance level is determined from the test image signals. Typical parameters of said distribution function are calculated and compared with sets of stored corresponding parameters derived in a similar manner from known reference sample images. A recognition signal is outputted to indicate the sample image which most closely resembles the test image as determined by said comparison, provided that the quantitatively determined differences between test and sample images are below a predetermined threshold value.