Observation of Digital and Film Images

Film and Si/CMOS digital cameras have been used for decades to record images of objects of interest. Film images produced by the Apollo program 70 mm Hasselblad cameras have been scanned and digitized to produce nominal 5700 x 5800 RGB images with 8-bit dynamic range per color channel. Film can be shown to be capable of  recording images with 12-bit dynamic range. Commercial DSLR cameras are capable of recording RGB images with 14-bit dynamic range for each color channel. Regarding recent NASA space missions, the Dawn spacecraft provides a good example. During its mission to Vesta and Ceres, it transmitted compressed image data that is processed to produce nominal 14-bit, 1024 x 1024 monochrome images. Displaying Dawn images on computer monitors with typical 8-bit intra-scene dynamic range constricts the number of available shades of gray from a possible 16,384 to 256. While display monitor loss is significant, it is not the weakest link in the observation channel.

The Human Visual System

The inter-scene dynamic range of human vision spans many decades, enabled by iris control and nonlinear retinal responsivity, allowing us to view terrestrial scenes in bright daylight and stars and planets at night.  However, our intra-scene dynamic range for a particular inter-scene vision adaptation is nominally 50 shades of gray, which corresponds to a minimum observable limiting (‘liminal”) contrast of 2% that can be expressed as a 6-bit digital value. Cobb, C.W., Moss, F.K. (1928), Connor, J.P., Ganoung, R.E. (1935), and Blackwell, H.R. (1946) performed vision measurements on human subjects that measured percent observable liminal contrast vs. reciprocal minimum resolvable angle and scene luminance.  Rose, A. (1973) prepared composite graphs of their data. The foregoing show that the human visual system liminal contrast is rarely below 2%.  In comparison, 14-bit digital images can record a minimum contrast of 6.1 x 10^-3 %. Our 6-Bit Vision is a severe mismatch for extracting information from 14-bit images; we are the weakest link in the observation channel.

Mitigating the Human Vision Mismatch –Dynamic Visibility Enhancement (dVE) Image Processing

Standard image processing tools available in many commercial software packages, such as background subtraction, linear transformations, the unsharp mask, and nonlinear transformations such as logarithmic, gamma, and histogram equalization, have been used to improve the visibility of image detail. However, they provide limited mitigation of our vision intrascene dynamic range mismatch.

Recognizing the foregoing limitations, Dr. Moran developed proprietary image processing algorithms and protocol, Dynamic Visibility Enhancement (dVE), that transforms high-intra-scene-dynamic-range image information for effective viewing by our 6-Bit Vision. It improves the visibility of information in images of typical scenes containing large regional variations of both brightness and contrast. dVE is nonlinear and spatially adaptive and can operate on both grayscale and color images.

dVE can also improve the visibility of spatial sensor and photoelectron shot noise in local image regions where it begins to dominate the visibility of the local image information; a reduction in the Local Signal-To-Noise Ratio (LSNR) The composite noise local visibility is a decreasing function of the LSNR.  An observer of dVE images can use this effect to identify image regions whose information content, and therefore usefulness, is LSNR limited.

The overall appearance of a dVE image when compared to the corresponding unprocessed image is one of increased texture that is directly attributable to the increased visibility of fine image detail that is below liminal contrast and normally invisible. We see and have become accustomed to a version of reality that has been smoothed by the liminal contrast thresholding of human vision.

In view of the foregoing it is clear that a) camera selection for critical observations should not only incorporate a sufficiently high number of pixels to resolve details of interest, but should also achieve the highest possible intra-scene dynamic range, b) image analysis should not be preceded by the truncation of high dynamic range images to 8 bits, and c) research imagery should always be subjected to advanced digital image processing such as dVE, otherwise critical image information may remain invisible.

dVE has been applied to images of terrestrial and space objects. The Dawn spacecraft framing camera images of the asteroid Vesta provide an example of the astonishing invisible image detail that can be revealed by applying dVE. These results raise serious questions regarding the possibility that critical information contained in thousands of current and past images of space and terrestrial objects remain invisible, locked from observation by our 6-Bit Vision.

References

Blackwell, H.R. (1946). Contrast thresholds of the human eye, Journal of the Optical Society of America, 36(11), pg. 624.

Cobb, C.W., Moss, F.K. (1928). The four variables of visual threshold. J. Franklin Institute, 205, pg. 831.

Connor, J.P., Ganoung, R.E. (1935). An experimental determination of visual thresholds at low values of illumination, Journal of the Optical Society of America, 25, pg. 287,

Rose, A. (1973). Vision: Human and Electronic, Plenum Press, New York.