While museums and archiving institutions are under pressure to make their document collections accessible to an increasing number of scientific researchers and other users, optimal preservation of paper-based documents requires storage in temperature and humidity conditions that preclude easy access and frequent handling. Very high quality digital imaging is seen by many museums as a means of providing document users simultaneous access to reproductions of close-to-original quality. However, there are few guidelines published concerning image quality. To develop a preliminary set of quality and compression standards, the Smithsonian's National Museum of Natural History requested that we examine, in detail, one of its collections of documents, namely its Fish Illustration Collection. Our aim was to develop specifications providing a level of detail that would satisfy demanding researchers and curators, yet not result in excessive storage and/or transmission time. Our examinations included visual analyses, spot metering with a digital colorimeter, and digital scanning at various sampling frequencies and levels of quantization of the originals and several types of photographic reproductions. The Museum's researchers and curators found that reproductions produced according to our specifications using carefully calibrated, high quality document scanners and computer monitors are very faithful to the originals.

Digitizing high-quality microscopic images and developing input/output technology for displaying those results is critical to tele-pathology in which pathological microscopic images are transferred to remote locations where they are diagnosed by specialists. This paper will discuss the results achieved by directly digitizing pathological microscopic images at a 2k by 2k resolution, and then using a super high definition imaging system to analyze their signals and evaluate compression performance. We will start off by digitizing samples that a pathologist will actually use in making a diagnosis, and then analyze their color distribution and spatial frequencies characteristics by comparing them to general images. This will make it apparent that such pathological images characteristically contain high spatial frequency in their chrominance components. We will also discuss the evaluation results of color differences for L.a.b space and compression ratios achieved when using JPEG to encode pathological images. We will also present a subjective evaluation of the influence sub- sampling of chrominance components has on image quality.

In order to obtain an extra high quality electronic imaging system, new important physical factors and system characteristics are considered and partly tested in addition to such traditional characteristics as the frequency characteristic and the signal to noise ratio. In particular, we consider the effect of the gama on the preservation of the quality, and the precision in the addition of R, G and B components in a display CRT, needed to maintain the high order global impression of an original image. The problem is discussed systematically both on the basis of measurements in the uniform lightness-chromaticity scale system and of high order psychophysical effects.

Electronic imaging may receive several different attributions in the domain of fine arts, and more precisely in the domain of paintings. Some of these attributions may appear sometimes conflicting. The most developed one is the reproduction of the main masterpieces under compact and universal forms allowing for a broad diffusion and display under many very different conditions. It favors efficient and reliable reproductions of color, and demands economic acquisition protocols and tries to mimic the popular slides which are available at every museum entrance.On the other side, digital imaging may be seen as the ultimate solution for a perennial archival of the paintings, or, under another light, as the safer way to allow many different experts to work simultaneously on the same document. It is clear that these two last applications will be drastically more demanding in the quality and the faithfulness of the reproduction they use. Our studies have been made with these goals in mind: i.e. to reproduce any painting with such a quality that the obtained documents may be considered as masters for successive degraded reproductions of the originals, or as sources for a reliable and complete information on the true painting.

Almost every major institution is contemplating some kind of digital-imaging project, but sources of information and experience are few. Technical aspects of digital imaging present libraries and archives with very difficult and complex choices. There are as yet no codified technical standards for image capture, display, and output, all of which affect the image quality, the cost, and, ultimately, the success or failure of the entire undertaking. Even if a vendor will provide a finished 'turnkey' system, an institution must understand the nature of the digital- imaging product they are buying. They must know how much image quality and functionality can be expected from it, both now and in the future. New tools must be provided for collection managers to make that possible. Beyond these purely technical issues, institutions must be able to relate the digital-image database system to the fundamental collection activities of access and preservation. Despite all the possibilities for manipulating digital images, image quality choices made when files are first created have the same 'finality' that they have in conventional photography. They will have a profound effect on project cost and the value of the final project to the users. Image quality requirements therefore have to be established carefully before a digitization project starts.

Digital photography was applied to the capture of images of the stained glass windows in the historic parish church in Fairford, Gloucestershire, England. Because of their size, the windows had to be photographed in 45 separate sections in order to capture all the detail present in the painting on the glass. The digital images of each section, approximately 3000 by 2300 pixels, were then mosaiced together in order to construct the very high resolution image needed for the complete window. A special backlight panel was constructed for the purpose, and techniques developed for minimizing the effects of reflected light and for calibrating the color of the images. Improvements in the technology for mounting and positioning the camera were identified as the most significant factors currently preventing the widespread adoption of this technology for virtual heritage applications.

This paper will describe a fine-art reproduction process that: captures painting information with high-resolution color photographs; scans the information into a 300 megabyte digital file; performs a 3D color calibration in a dedicated hardware color-transform circuit; makes a master positive color transparency and makes a reproduction on polaroid color print film. The master transparency can be used to expose a large number of images. This combines the efficiency of instant photography with the color fidelity of digital color transforms.

While a losslessly compressed facsimile image might require 20,000 bytes of storage, a losslessly compressed color high resolution scan of the same sized document might require 200,000,000 bytes of storage. This factor of 10,000 in the image size necessitates more than just better compression, it requires a change in viewpoint about compression. A compression system for high quality images must provide a way to access only the required data rather than decompressing all the data and then selecting the desired portion. Furthermore, a high quality image compression system should be able to provide the best possible images for output devices which as of yet have not been manufactured. Finally, a high quality compression system should allow decompression and recompression without continual degradation of the image. This paper describes technologies including a reversible color transform, a reversible wavelet transform, a doubly embedded context mode, and a 'parseable' file format, which work together to provide solutions for high quality imaging needs.

In recent work, we have examined the performance of wavelet coders using a perceptually relevant image quality metric, the picture quality scale (PQS). In that study, we considered some of the design options available with respect to choice of wavelet basis, quantizer, and method for error- free encoding of the quantized coefficients, including the EZW methodology. A specific combination of these design options provides the best trade off between performance and PQS quality. Here, we extend this comparison by evaluating the performance of JPEG and the previously chosen optimal wavelet scheme, focusing principally on the high quality range.

We study the use of spectral deltas for lossless multispectral image compression. Spectral deltas are differences between prediction errors. When bands are correlated, the prediction errors between the two bands are similar, thus the difference results in a smaller value. Building upon earlier work, we examine methods of detecting correlations and harnessing current advances in lossless still image compression. THe result is an algorithm that works well over a broad set of test images.

We consider here the compression of still color image with very low distortion from the human eye point of view. The basic idea in this work is to take into account the variations of human eye/brain spatial resolution with color. The most natural way for an image processing researcher to perform such a scheme is to use a multiresolution analysis of the image to be coded before quantization and coding. Previous experiences connected with still grey value image compression/decompression scheme design have shown that the wavelet transform, Mallats algorithm is a very efficient method for this purpose, particularly if real time implementation is under consideration. Hence we present in this paper a wavelet transform algorithm for color image and we show how and with what performances the transformed image can be altered and reduced. We show that a quasi lossless compression/decompression scheme can be easily obtained with compression ratio up to 1:10. The results obtained after a large series of testes based on psychovisual estimations rather than on pure PSNR evaluation are in good accordance with the assumed properties of the human visual perceptive system.

This paper clarifies two traditional image size reduction and enlargement methods, projection-method (PM) and linear- interpolation (LI), in terms of multi-rate signal processing. We have proved that both size conversion methods of scaling ratio of N/M can be implemented with the same structure, i.e., zero-hold enlargement by 1:n, average filtering of M-taps for PM or N-taps for LI, and M:1 down- sampling for both methods. Such implementation enables objective evaluations of PM and LI under the same manner that evaluates other size conversion methods originally developed with digital filtering. The empirical matter that PM provides better image quality in reduction than LI and the opposite behavior in enlargement is confirmed under multi-rate signal processing. We have developed a unified method of PM and LI by introducing selective filter-length. N-taps for enlargement and M-taps for reduction are used to produce better or equal image quality at any scaling ratio of N/M than PM or LI alone does. We also show that minute filter-length adjustment of the unified method can control image quality more delicately; a slight decrease of the filter-length provides sharper images and a slight increase provides smoother images.

This paper introduces a new approach to deblocking of JPEG compressed images using over-complete wavelet representations. By exploiting cross-scale correlations among wavelet coefficients, edge information in the JPEG compressed images is extracted and protected, while blocky noise in the smooth background regions is smoothed out in the wavelet domain. Compared with the iterative methods reported in the literature, our simple wavelet-based method has much lower computational complexity, yet it is capable of achieving the same PSNR improvement as the best iterative method, and giving visually very pleasing images as well.

Typical objective methods for quantifying image quality, as part of evaluating coder performance, are obtained by computing a single or several numbers as a function of the difference image between the original and coded images. Pre- processing images prior to encoding can remove noise, or unimportant detail, and thus improve the overall performance of the coder.However, the error image obtained with the pre- processed image as a reference is substantially different than the one obtained if the original. This paper addresses the issue of combining the changes in the image due to pre- processing and the degradation due to encoding. The objective is to obtain global quality measures that quantify the value of pre-processing for image coding.

It is generally accepted that a RGB color image can be easily encoded by using a gray-scale compression technique on each of the three color planes. Such an approach, however, fails to take into account correlations existing between color planes and perceptual factors. We evaluated several linear and non-linear color spaces, some introduced by the CIE, compressed with the vector quantization technique for minimum perceptual distortion. To study these distortions, we measured contrast and luminance of the video framebuffer, to precisely control color. We then obtained psychophysical judgements to measure how well these methods work to minimize perceptual distortion in a variety of color space.

The European Commission-funded MARC project ended in April 1996, with the publication of Flemish Baroque Painting, Masterpieces of the Alte Pinakothek, Muenchen (Hirmer 1996). To the best of our knowledge, this is the world's first all- digital colorimetric art catalogue. This paper will briefly introduce the MARC camera and the MARC printing technology, and then present a critical evaluation of the final book. The application of MARC results since the end of the project will be covered, and related EC imaging projects surveyed.

We describe computational experiments to predict the perceived quality of multilevel halftone images. Our computations were based on a spatial color difference metric, S-CIELAB, that is an extension of CIELAB, a widely used industry standard. CIELAB predicts the discriminability of large uniform color patches. S-CIELAB includes a pre- processing stage that accounts for certain aspects of the spatial sensitivity to different colors. From simulations applied to multilevel halftone images, we found that (a) for grayscale image, L-spacing of the halftone levels results in better halftone quality than linear-spacing of the levels; (b) for color images, increasing the number of halftone levels for magenta ink results in the most significant improvement in halftone quality. Increasing the number of halftone levels of the yellow ink resulted in the least improvement.

We propose a new method to improve the design of electro- optical imaging system using an end-to-end model of the imaging systems and a combination of image quality criteria. Firstly, we used an imaging systems simulator to produce an output image, which is the distorted version of the input scene. Secondly, we calculate an objective quotation for the quality of the output images corresponding to several systems configurations, and compare the results with the human evaluation of image quality. It allows us to calibrate our imaging systems quality measure (ISQM). Finally, the ISQM is used as a tool to improve system design without any human observer evaluation.

GPA is an expression that describes how the number of dots/inch and the number of graylevels/dot tradeoff in determining the number of graylevels per area (GPA). The metric is based on the assumption that anything that falls within a visual angle of approximately 3 minutes of arc will be spatially-integrated by the optical blur properties of our eyes.

In this paper we present a metric that predicts differences in perceptual sharpness from physical parameters extracted from CCD camera recordings of sine-grating and step-profile test images. Sharpness differences were induced by using five different high-end television sets to display natural color images as well as by using blurred versions of these images. The metric is a two-step model that uses the perceptual strengths of three sharpness-related attributes on an intermediate level between the physical parameters and perceptual sharpness. The data obtained from category scaling experiments indicate that perceptual strength of sharpness is a linear combination of the perceptual strengths of the attributes detail rendering and contour rendering. It was found that differences in detail rendering correlate with differences in modulation depth of the sine gratings and that differences in contour rendering can be explained from the slope and overshoot parameters of the step-profiles.