Carol L. Novak

Anatomy of a color histogram

It is shown that the color histogram has an even closer relationship to scene properties than has been previously described. Color histograms have identifiable features that relate in a precise mathematical way to scene properties. Object color and illumination color are the most obvious properties that are related to color distribution, and their extraction has already been described. It is shown here that the histogram of color variation may be further exploited to relate its shape to surface roughness and imaging geometry. An understanding of these features allows an improved estimate of illumination color and object color to be made.<<ETX>>

Obtaining accurate color images for machine-vision research

Algorithms for color machine vision rely upon an explicit or implicit model of how color images are formed. These models usually include an idealized description of how camera systems measure color. Unfortunately, relying on these idealizations can cause serious errors in the performance of algorithms. In this paper, we describe several problems that occur in real color images and how they can be dealt with. A common problem is that many cameras do not have a linear response. Failure to correct this problem will not only cause hue shifts, but will cause problems in any algorithms that rely on the linear assumption. We show how this effect may be measured and corrected. More insidious are the problems of color clipping and blooming, since they can cause sudden changes in the apparent hue. Although it is not possible to correct the problem, we show how it may be detected so that affected measurements may be discounted. Most models assume that integration is performed over the visible spectrum, but many cameras are sensitive to other regions as well. Failure to compensate for this can lead to unexpected results. Also the camera’s varying spectral sensitivity within the visible spectrum will cause some measurements to be more prone to noise than others. We show how the use of a computer- controlled lens can compensate for this problem. Chromatic aberration in typical video camera lenses can cause hue changes, and we show how the further use of computer-controlled lens can compensate for this problem. Using these methods to detect and/or correct problems, we can obtain the accuracy needed for applying physics- based methods to actual color images.

Method for estimating scene parameters from color histograms

One of the key tools in applying physics-based models to machine vision has been the analysis of color In the mid-1980’s it was recognized that the color histogram for a single inhomogeneous surface histograms. with highlights will have a planar distribution in color space. It has since been shown that the colors do not in a plane but form clusters at specific points. Physics-based models of reflection predict that fall randomly shape of the histogram is related not only to the illumination color and the object color but also to such the properties as surface roughness and imaging geometry. We present an algorithm for analyzing color noncolor histograms that yields estimates of surface roughness, phase angle between the camera and the light source, and illumination intensity. These three scene parameters are related to three histogram measurements. However, the relationship is complex and cannot be solved analytically. Therefore we developed a method for estimating these properties that is based on interpolation between histograms that come from images of known scene properties. We present tests of our algorithm on simulated data, and the results compare well with the known simulation parameters. We also test our method on real images, and the results compare favorably with the actual parameters estimated by other means. Our method for estimating scene properties is very fast and requires only a single color image.

Physics-based models for early vision by machine

Research in early (low-level) vision, tooth for machines and humans, hasntraditionally been based on the study of idealized images or image patches such as step edges, gratings, flat fields, and Mondrians. Real images, however, exhibit much richer and more complex structure, whose nature is determined by the physical and geometric properties of illumination, reflection, and imaging. By understanding these physical relationships, a new kind of early vision analysis is made possible. In this paper, we describe a progression of models of imaging physics that present a much more complex and realistic set of image relationships than are commonly assumed in early vision research. We begin with the Dichromatic Reflection Model, which describes how highlights and color are related in images of dielectrics such as plastic and painted surfaces. This gives rise to a mathematical relationship in color space to separate highlights from object color. Perceptions of shape, surface roughness/texture, and illumination color are readily derived from this analysis. We next show how this can be extended to images of several objects, by deriving local color variation relationships from the basic model. The resulting method for color image analysis has been successfully applied in machine vision experiments in our laboratory. Yet another extension is to account for inter-reflection among multiple objects.nWe have derived a simple model of color inter-reflection that accounts for the basic phenomena, and report on this model and how we are applying it. In general, the concept of illumination for vision should account for the entire illumination environment, rather than being restricted to a single light source. This work shows that the basic physical relationships give rise to very structured image properties, which can be a more valid basis for early vision than the traditional idealized image patterns.