Low bitrate coding for storage of ID photos

The application of suboptimal runlength codes to two-tone ID photos is discussed. Standard and vertical predictive run-length-coding techniques are examined. Huffman coding of the results is reviewed. Computation of thresholds, edge detection, and pattern recognition are employed to enhance the images. A 256×256-pixel picture, quantized to 64 grey levels, was scaled down to two levels, employing a histogram averaging algorithm to determine the most effective threshold. Suboptimal vertical adaptive coding was implemented to minimize digital storage space. The bit rate was reduced to 0.145 b/pel on average with a deviation of σ=0.01722. Image enhancement was artificially produced by superimposing the eyes from the edge detection sequence onto the two-tone picture. Pattern recognition was employed to detect the nose, thus creating a lower bound for the algorithm searching for the eyeline while defining a natural axial line through the facial image