Dynamic range compression of video-phone images by simulating additional diffuse scene illumination in their video signals

The mismatch between dynamic ranges of intensities in non—uniformly illuminated videophonescenes and of intensities reproducable at display devices is addressed. Displayed portrait images with no details visible in some facial regions result from this mismatch and lead in turn to a reduced impression of telepresence.A novel image processing approach for dynamic range compression of videophone portrait imagesis presented in this paper. Unlike conventional methods, it provides compressed images with a mostnatural appearance. The basic idea is to simulate a more uniform scene illumination in the video signalsof videophone portrait images. This requires a determination of the scene's reflectance function. Anextension of the classical lightness approach of Land and McCann [5] to curved surfaces is proposedfor this task.Keywords: dynamic range compression, videotelephony, lightness algorithm 1 INTRODUCTION In the near future visual telecommunication will become common practice via videophone sets fromhomes, offices etc. Like with today's plain old telephones, users will want these terminals to work intheir existing environments. They would not accept specially designed videophone studios. However, inordinary domestic or working environments one is faced with various lighting conditions (bright sunlight,dark daylight, or artificial light) [8]. Specifically, if direct sunlight is falling in, the illumination of thescenes can be highly non—uniform. In these cases, intensity variations between those portions of a faceturned towards and those turned away from the sun may easily exceed ratios of 500 : 1. Electronicallyreproduced images of such scenes show obscured regions because of the limited dynamic range of commondisplay devices1. Figure 1 illustrates this for a typical videophone portrait image.Using conventional dynamic range compression techniques such as histogram transformation [3] orhomomorphic filtering [9], details may become visible in otherwise obscured regions of the image. But,these techniques generate compressed images which tend to appear unrealistic2 {2]. Thus, they are unsuit-able to resolve dynamic range problems in videotelephony or videoconferencing, since these services areaimed at giving the impression of a communication situation similar to that of real face—to—face meetings