Ralph E. Jacobson

An evaluation of image quality metrics

Many metrics have been proposed, over a period of thirty years, which are able to predict image quality in the form of single numbers. These metrics reduce a number of physical measurements to single values for predictions of perceptions of the quality of many different types of images from statistical evaluations by panels of observers. One problem concerned with the determination of any metric is which eye modulation transfer function (contrast sensitivity function, CSF) to use. It is only since 1992 that a physical model for determining the CSF of the eye viewing conditions. This paper reviews image quality metrics for sharpness and noise both independently and in combination

Practical camera characterization for colour measurement

Abstract This paper reviews the major issues involved in the use of digital cameras to derive the CIE X, Y and Z tristimulus values of objects in real scenes. Both practical and theoretical investigations have been carried out to gain experience in this specialized field of imaging that is finding application in, for example, machine vision, product quality assessment by panel observation and digital archiving of art objects. The practicalities of camera characterization described include: lighting—spectral power and uniformity; test target—choice and number of colours; camera signal processing—linear or gamma corrected; colour analysis—filter transmittance and infrared filtration; characterization method—linear or higher order; quality measure—CIELAB, CMC, CIE94 colour difference; quality statistic—mean, median, etc. The potential impact of these parameters is discussed by the use of a computer model and from practical experience. It is shown, as might be expected, that the choice of colour separation filtration is the most sensitive variable. If a ‘colour’ camera is used, then it needs to be carefully selected; a more adaptable choice, however, may be a monochrome camera with external filters. In addition, the illumination uniformity of the test target used for characterization is shown to be important: that it is never perfectly uniform must be considered in the characterization process. With careful selection of system components, a median value of less than 1.0 CIELAB colour difference can be obtained between independently measured colorimetry and that calculated from the camera output.

Minimum Perceptible Differences in the Colour Reproduction of Photographic Prints

AbstractThe aim of this work was to investigate observer perceptibility of small colour differences in the colour reproduction of reflection prints produced using a negative/positive photographic system. Comparison of the experimental results, in the form of hue-oriented and non hue- oriented ellipses, was made with MacAdam ellipses, CIELAB unit ellipses and ellipses derived from the CMC (l:c) colour difference formula, in the 1976 CIELAB colour space. It was found that the CMC (l:c) ellipses fitted the hue-oriented ellipses more closely than the CIELAB unit el-lipses, there being no significant differences between the hue-oriented and the non hue-oriented ellipses. There was, however, a significant difference between MacAdam ellipses and the non-hue oriented experimental ellipses. It was therefore concluded that the CMC (l:c) colour difference formula could be used to predict the tolerance ellipse for any colour centre, under the conditions used in this work. The minimum perceptible difference was found ...

Food colour appearance judged using images on a computer display

Abstract Pixel-based images of food products are prepared and systematically perturbed in colour and lightness, and the changes quantified. Observations are reported in terms of CIELAB and pixel value, with colour differences corresponding to acceptability and perceptibility thresholds for each subject and each perturbation applied. Conclusions are drawn concerning the usefulness of digital imagery for product panel assessment, the relative importance of colour and lightness variations of the food products investigated and the applicability of a mathematical function to the results of the visual assessments carried out. The relationship of acceptability to imperceptibility of image perturbations is quantified.

Assessment of a CRT display system

The use of cathode ray tube (CRT) monitors and graphics cards for the display of pictorial images is common. Certain applications require the characteristics of the display device or medium to be accurately understood. One such application is the use of CRT based displays in experiments involving subjective analysis of viewed images. In this paper we describe a series of experiments performed to characterize a display device used in our laboratories prior to its use in a subjective assessment experiment. The tone reproduction characteristics are studied using transfer functions and gamma. Colour reproduction is studied in temporal and spatial dimensions. Finally, spatial effects are characterized by use of modulation transfer function (MTF) analysis.

The Application of Colour Reproduction Indices to Photographic Reflection Prints

AbstractA possible correlation between colour reproduction indices derived from a recent model of colour vision, and subjective rankings of photographic reflection prints, was sought. Sets of reflection prints consisting of a normal print and a colour ring-around were produced for four scenes. Observers placed in rank order the prints in each set according to the quality of colour reproduction. Ten observations were made for each set of prints; viewing was made in a viewing booth using standard daylight conditions. From colorimetric measurements of the Macbeth ColorChecker Chart set of prints, colour reproduction indices were calculated. A reasonable correlation between the implied rank order derived from the colour reproduction indices and experimental rank order was found.

Just noticeable gamma differences and acceptability of sRGB images displayed on a CRT monitor

Abstract The standard RGB colour space (sRGB) has been proposed as a means for obtaining accurate reproduction of colour and tone for images displayed across the Internet, provided that they are viewed under the reference display and viewing conditions defined in the standard. It has been found, however, that typical display and viewing conditions when accessing online images vary and therefore deviate from the reference sRGB conditions. One of the parameters that may affect the perceived quality of online images is the gamma setting of the display. In this work psychophysical experiments were conducted to determine the imperceptibility and acceptability of gamma differences of sRGB images when they are viewed on cathode ray tube displays. These experiments were carried out under both controlled and uncontrolled display and viewing conditions. The results of these experiments are presented and discussed, including the estimated points of subjective equality and the just noticeable difference of gamma values.

Perceived colour differences in displayed colours Part 1: hard copy to soft copy matching

Abstract An investigation is reported into visual comparisons between object colours and simulations on a cathode ray tube (CRT) monitor display. Experiments are described in which observers viewed a bipartite field in which two coloured areas were separated by a narrow division and viewed against an unlit grey background. One coloured area comprised a surface colour patch illuminated using a light source which was a close colorimetric match to CIE Illuminant D65. The second area was simply an aperture held against the faceplate of the computer CRT monitor, operated with a white point corresponding again to CIE Illuminant D65. The illumination of the reflection patch was arranged so that there was a visual match between a white patch and the monitor white. The monitor patch was controlled in appearance from the keyboard and was specified in terms of red, green and blue pixel values. Observers were asked to match the illuminated reflection patch by adjusting the (unseen) specification of the monitor patch. Separate experiments concerning the ability of observers to detect colour perturbations of matches are also described. Conclusions are drawn as to the concordance of repeated colour matches by individual observers, and between all the observations. The magnitudes of perceptible colour differences are also evaluated.

Observer responses to colour perturbations in reflection prints

Abstract The results of investigations into observer response to perturbations of colour prints with known CIELAB colour differences from a standard reproduction are assessed. The responses in terms of acceptability and perceptibility of colour differences are related to percentage imperceptibility, being defined as 100—perceptibility. The acceptability and imperceptibility of colour reproduction are evaluated as functions of CIELAB colour difference, giving good fits to a single simple mathematical function but with differing parameters. A near-constant ratio of acceptable to imperceptible colour difference is found to apply over the entire range of observer responses.

Determination and Enhancement of Image Contrast Using the Exposure Density Concept

AbstractColour filters arc commonly used in monochrome photography for altering the relative lightness of different coloured parts of the original scene when recorded in the image. However, the choice of filter is not straightforward where it is necessary to enhance contrast between closely similar colours. The concept ofexposu re density can be used to quantify the effect on the image of different filters, combined with the effect of the spectral properties of all the system’s parameters such as the light source, response of the sensor and transmittance of the camera lens. This method can be applied to any other imaging system and is not restricted to visible radiation. Calculations of predicted exposure density difference, or system exposure response difference (SERD) for a digital image, provide the basis for the selection of the optimum filters and other system parameters for maximizing image contrast in the recording of closely similar colours. 82