During the first day following the Viking 1 early morning landing we were preoccupied with analysis and release of the first two pictures which, in quality and content, had greatly exceeded our expectations. Almost 24 hr later, when the Orbiter once more passed over the Lander, another picture was relayed to Earth, this one in color.
In contrast to the attention we lavished on planning of the first two black and white pictures, we were dismally unprepared to reconstruct and analyze the first color picture. In a general way we understood that thorough preflight calibration of the camera's spectral sensitivity was mandatory. We also knew that we would need software programs that efficiently transformed the raw data. What we failed to appreciate were the many subtle problems which, uncorrected, could produce major changes in color. Furthermore, we had no intimation of the immediate and widespread public interest in the first color products-for example, intuitively corrected color images were shown on television within 30 min following receipt of data on Earth. Although we struggled to delay the deadline, we were obliged to release the first color prints within 8 hr after receipt of data.
As previously mentioned, there are three sensors with blue, green, and red filters in the focal plane of the camera. These record the radiance of the scene in blue, green, and red light. However, the multilayer interference filters used in the cameras (simpler absorptive emulsion layers would have been degraded by preflight heat sterilization) have very irregular spectral response. The blue channel, for example, responds slightly but significantly to infrared light. The extraneous parts of the signal must be subtracted, so that the absolute radiances at three specific wavelengths in the blue, green, and red are represented. A color print is produced by exposing conventional color film to separately modulated beams of blue, green, and red laser light, scanning the film with the same geometry employed in the camera.
Preflight calibration of the camera thoroughly characterized the sensitivity of each sensor filter combination. Qualitative tests indicated that simple normalization of the voltages for the three color channels-disregarding spectra leaks and displacements-was sufficient to produce reasonable color images. In making that judgment our attention was generally directed to saturated colors in the natural scene and test target.
When the first color data from Marc were received on Earth, we immediately used the same normalization techniques to calibrate the image. The result was surprising and disquieting. The entire scene, ground and atmosphere alike, was bathed in a reddish glow. Unwilling to commit ourselves publicly to this provocative display, we adjusted the parameters in the calibration program until the sky came out a neutral gray. At the same time, rocks and soil showed good contrast; the colors seemed reasonable. This was the picture released eight hours after receipt of the data. But to our chagrin the sky took on a bluish hue during reconstruction and photoreproduction. The media representatives were delighted with the Earth like colors of the scene.
Meanwhile, continued analysis supported the reality of an orangeish tint throughout the scene, the atmospheric color resulting from small suspended soil particles. Several days after the first release, we distributed a second version, this time with the sky reddish. Predictably, newspaper headlines of "Martian sky turns from blue to red" were followed by accounts of scientific fallibility. We smiled painfully when reporters asked us if the sky would turn green in a subsequent version.
Our work over the past year has demonstrated that, even though the sky will not turn green, there will be nevertheless a long series of color images, each better than-or at least different from-its predecessor. The initial images were unnaturally saturated, or rich in chrome. Reducing the saturation makes the scene appear more drab. The reader should keep in mind that color is defined according to three properties: hue, the characteristic wavelength of the pigmenting material; saturation, the amount of pigmenting material; and brightness, the admixture of pigmenting material and gray background. Most persons equate apparent color with hue. However, changes in brightness and saturation can produce images which appear to be much different in hue.
More important than correcting for saturation was accounting for spectral "out of band" filter transmittance. In order to do this it is necessary to examine, for each picture element, reflectances in all six bands, three in visual color and three in near infrared. The amount of off axis radiance for each channel is estimated by considering the radiance in the five other channels and developing an integrated spectral solution that satisfactorily accounts for the measured reflectances in all six channels.
The reader may be thinking that problems of color reconstruction largely could have been circumvented by mounting on the spacecraft targets of known color. In fact, such an array of color chips was mounted and has been imaged. We have discovered that accurate color reconstruction of this target is necessary, but not sufficient for certification of good color in the natural scene. Errors are not particularly noticeable for saturated colors but become significant at low values of brightness and saturation.
In summary, the color of the martian scene, perceived by the necessarily abnormal eyes of Viking, is elusive. In response to the inevitable question: "Is that exactly how it would look if I were standing on Mars?" a qualified "yes" is in order.