This PDF file contains the editorial “Editorial (October 1995)” for JEI Vol. 4 Issue 04

A visualization procedure for the 3-D histogram of color images is presented. The procedure assumes that the histogram is available as a table that associates to a pixel color the number of its appearances in the image. The procedure runs for the RGB, YMC, HSV, HSL, L*a*b*, and L*u*v* color spaces and it is easily extendable to other color spaces if the analytical form of color transformations is available. Each histogram value is represented in the color space as a colored ball, in a position corresponding to the place of the color in the space. A simple drawing procedure is used instead of more complicated 3-D rendering techniques. The 3-D histogram visualization offers a clear and intuitive representation of the color distribution of an image. The procedure is applied to derive a clusterization technique for color classification and visualize its resuIts, to display comparatively the gamut of different color devices, and to detect the misalignment of the Rc3B planes of a color image. Diagrams illustrating the visualization procedure are presented for each application.

Color-appearance models are used to relate chromatic stimuli viewed under one set of viewing and illuminating conditions to a differing set such that when each stimulus is viewed in its respective conditions, the stimuli match in color appearance. These models assume the observer has a steady-state adaptation to each condition. In practice, observers often view stimuli under mixed adaptation; this could occur when viewing CRT and reflection-print stimuli simultaneously. A visual experiment was performed to determine whether the RLAB color-appearance model could be used successfully to generate reflection prints that match the appearance of the CRT when viewed under mixed states of adaptation and in turn as stand-alone images viewed under a single state of adaptation. Sixteen observers viewed four pictorial images displayed on a D₆₅ balanced CRT display in a room lit with cool-white fluorescent luminaries. The RLAB color-appearance model was used to calculatecorresponding images where the observer's state of chromatic adaptation was assumed to be one of the following: adaptation to each device condition, a single adaptation at the midpoint of the two device conditions, adaptation to the CRT condition and a print adaptation shifted 25% toward the CRT condition, adaptation to the print condition and a CRT adaptation shifted 25% toward the print condition, and a CRT condition shifted 25% toward the print condition and a print condition shifted 25% toward the CRT condition. Each condition was compared pairwise and Thurstone's law of comparative judgments was used to calculate interval scales of quality. Observers first judged the reflection prints adjacent to the CRT display selecting the image closest in color appearance to the CRT image; they also categorized the closest image as "acceptable, " "marginally acceptable," or "not acceptable." The images were again scaled except the display was turned off; this determined the best standalone color reproduction.

A desktop drum scanner was colorimetrically characterized to an average CIELAB error of less than unity for Kodak Ektachrome transparencies and Ektacolor paper, and Fuji Photo Film Fujichrome transparencies and Fujicolor paper. Independent verification on spectrally similar materials yielded an average ΔE*_ab error of less than 2.1. The image formation of each medium was first modeled using either Beer-Bouguer or Kubelka-Munk theories and eigenvector analysis. Scanner digital values were then empirically related to dye concentrations using polynomial step-wise multiple-linear regression. These empirical matrices were required because the scanner's system spectral responsivities had excessively wide bandwidths. From these estimated dye concentrations, either a spectral transmittance or spectral reflectance factor was calculated from an a priori spectral analysis of each medium. The spectral estimates can be used to calculate tristimulus values for any illuminant and obsetver of interest. The methods used in this research are based on historical methods commonly used in photographic science.

The Joint Photographic Experts Group's image compression algorithm has been shown to provide a very efficient and powerful method of compressing images. However, there is little substantive information about which color space should be utilized when implementing the JPEG algorithm. Currently, the JPEG algorithm is set up for use with any three-component color space. The objective of this research is to determine whether or not the color space selected will significantly improve the image compression. The RGB, XYZ, YIQ, CIELAB, CIELUV, and CIELAB LCh color spaces were examined and compared. Both numerical measures and psycho-physical techniques were used to assess the results. The final resuLts indicate that the device space, RGB, is the worst color space to compress images. In comparison, the nonlinear transforms of the device space, CIELAB and CIELUV, are the best color spaces to compress images. The XYZ, YIQ, and CIELAB LCh color spaces resulted in intermediate levels of compression.

A clustering algorithm for analyzing and partitioning the color images of natural scenes is described. The proposed method operates in the 1976 CIE (L*,a*,b*) uniform color coordinate system. It detects image clusters in some circular-cylindrical decision elements of the color space. This estimates the clusters' color distributions without imposing any constraints on their forms. Surfaces of the decision elements are formed with constant lightness and constant chromaticity loci. Each surface is obtained using only 1-D histograms of the L*,H°,C* cylindrical coordinates of the image data or the extracted feature vector. The Fisher linear discriminant method is then used to project simultaneously the detected color clusters onto a line for 1-D thresholding. This permits utilization of all the color properties for segmentation, and inherently recognizes their respective cross correlation. In this respect, the proposed algorithm also differs from the multiple histogram-based thresholding schemes in that it generates more reliable gross segmentation results.

Two experiments for evaluating psychophysical distortion metrics in Joint Photographic Experts Group (JPEG) encoded images are described. The first is a threshold experiment, in which subjects determined the bit rate or level of distortion at which distortion was just noticeable. The second is a suprathreshold experiment in which subjects ranked image blocks according to perceived distortion. The results of these experiments were used to determine the predictive value of a number of computed image distortion metrics. It was found that mean-square-error is not a good predictor of distortion thresholds or suprathreshold perceived distortion. Some simple pointwise measures were in good agreement with psycho-physical data; other more computationally intensive metrics involving spatial properties of the human visual system gave mixed results. It was determined that mean intensity, which is not accounted for in the JPEG algorithm, plays a significant role in perceived distortion.

An optimal model-based neural evaluation algorithm and an iterative gradient optimization algorithm used in image restoration and statistical filtering are presented. The relationship between the two algorithms is studied. We show that under the symmetric positive-definite condition, a condition easily satisfied in restoration and filtering, intra-pixel sequential processing (IPSP) of model-based neuron evaluation is equivalent to the iterative gradient optimization algorithm. We also show that although both methods provide feasible solutions to fast spatial domain implementation of restoration and filtering techniques, the iterative gradient algorithm is in fact more efficient than the IPSP neuron evaluation method. Visual examples are provided to compare the performance of the two approaches.

We present a viewing model that is appropriate for several types of displays used in virtual environment systems, including head-mounted displays and head-tracked stationary displays. The model accounts for arbitrary size, placement, and orientation of the display images, and thus is suitable for various display designs. We provide algorithms for calculating stereoscopic viewing and projection matrices. The tracking algorithm models the position and orientation of the tracker's emitter and the displacement between the sensor and the user's eyes. The algorithms are presented as parameterized homogeneous transforms. We also discuss features that can be used to avoid accommodation/convergence conflicts. The advantages of this viewing model and algorithm are the elimination of possible vertical parallax, an undistorted perception of depth, and reduction of eye fatigue due to excessive parallax. All of these factors contribute to improved comfort and utility for the operator.

This PDF file contains the editorial “Book Rvw: Real-time Computer Vision," edited by Christopher M. Brown and Demetri Terzopoulos for JEI Vol. 4 Issue 04