Digitization of printed illustrations for archive purposes is not always a simple procedure, due to the varied nature of illustration processes. For example, colored engravings were once the predominant type of book illustrations. Due to the specific printing process they consist of regions of both very fine lines (high-frequency) and almost uniformly colored regions (low-frequency). In order to keep the compression rate high, compression algorithms tend to destroy the fine lines that are the essence of this type of image. However, the images compressed using the modified mixed raster content (MRC) compression scheme showed exceptional fidelity to their prototypes at low bit-rates when compared to alternate compression methods such as JPEG and JPEG2000. This paper presents the simple segmentation algorithm for fast separation of engraved and colored layers, as a part of the image processing process necessary for MRC scheme implementation. This is demonstrated on the case study of the William Blake Archive.

Digitization of printed illustrations for archive purposes is not always a simple procedure. For example, colored engravings (a common book-illustration type of the XVIII and XIX century) consist of regions of both very fine lines and almost uniformly colored regions. In order to keep the compression rate high, compression algorithms tend to destroy fine lines (the high-frequency content) that are the essence of this type of image. However, the images compressed using the modified mixed raster content (mMRC) compression scheme retain the lines and show exceptional fidelity to their prototypes at low bit-rates when compared to alternate compression methods such as JPEG and JPEG2000. This paper presents the simple algorithm for data-filling of the MRC background layer, that improves the compression results, and - at the same time - enhances the reconstructed image. This is demonstrated on the case study of the colored engravings from the William Blake Archive.

The two-channel picture coding system previously reported is improved by adaptive companding and extended to component color coding. The basic system divides the spectrum into lowand high-frequency spatial components. The former are coarsely sampled and finely quantized, while the latter are finely sampled and coarsely quantized using a companded, randomized quantizer. The adaptive version uses a companding factor, larger than one, which multiplies the highs before coding and divides the result after decoding. This substantially raises the signal-to-noise ratio, which is calculated for both the adaptive and nonadaptive cases and confirmed by computer simulation. The color version uses an achromatic highs channel as in the basic monochrome system, plus a three-color lows channel coded in a perceptually uniform color space. With 3.5-4.3 bits/pel for 512 × 512 monochrome and color still images, the nonstatistical version produces images free of visible defects under normal viewing conditions. Projections are made for system parameters capable of achieving the very high quality now being proposed for production component coding systems, but with substantial reduction in channel capacity.

The two-channel system previously reported has been implemented in hardware using system parameters appropriate to consumer television. The basic system divides the spectrum into lowand high-frequency spatial components. The lows are coarsely sampled and finely quantized and the highs finely sampled and coarsely quantized using a companded, randomized quantizer. The purpose of this experiment was to demonstrate the potential of the system for low-cost practical application, to study the effect of the character of the randomizing noise, and to ascertain that there were no deleterious effects due to interlace, motion, or input noise. Theoretical noise calculations were qualitatively confirmed.