Making Informed Judgments Using Digital Images
Written by David “Ski” Witzke   

We have all heard endless clichés about photo realism, such as, A picture is worth a thousand words, or Photos don’t lie. But in the courtroom, a very different image often emerges because of clichés like, Can juries really believe what they see?, Digital images can be manipulated easily, or Digital images are not photographs. These clichés—and many more like them—can be very persuasive and invoke an emotional or cognitive response.

See this article in its original format in the Digital Edition!


Advertisement

Special procedures must be followed with digital images so they can be validated as a true and correct copy that accurately and reliably depicts the subject or scene. In addition, the forensic analyst as well as the juror must be able to visualize (or analyze) the content of the image, and to reach the “correct” conclusion.

The reason why lawyers (and others) often take the position that digital images are not photographs is because they want to instill the idea that jurors cannot believe what they see. This confusion exists because digital cameras capture images using an electronic sensor used to record light values such as a CCD (charge-coupled device) or CMOS (complementary metal-oxide semiconductor). These light values are then converted into numerical values that can be displayed or printed as an image, can be reproduced accurately and reliably, and do not degrade over time.

Traditional film cameras utilize a light-sensitive emulsion containing silver halide crystals suspended in a gelatin coating on films of cellulose acetate or cellulose nitrate. Once the film is exposed and then developed, a negative is created. The negative then becomes the “original” that can be reproduced accurately and reliably, but suffers the effects of aging over time.

Most people forget that the quality (clarity and contrast) of photographs (prints) produced from negatives are the interpretation of the person making the print, who is often not the photographer who actually took the picture. Anyone who has ever had pictures developed from film knows that there are many things that can go wrong, and the images do not always appear like you thought they would or should. Photographers who develop and print their own negatives get the most realistic (accurate) photographs, which is most often the case with digital imaging.

Television programs like Brain Games have featured optical illusions that demonstrate issues effecting visualization as well as believability or reliability. So whether it was a film image or a digital image, the same issues exist regarding “visualization” or “interpretation” of the image.

Crime scene technicians, forensic analysts and examiners, investigators, prosecutors, defense attorneys, jurors, or anyone who relies on “pictures” within the criminal justice community must make conclusions based upon their “interpretation” of a visual representation (input). This process must account for biases and other factors that can influence their decision, such as the proper evaluation of ridge detail that may appear to be inverted in photographed images of impressions on evidence developed using cyanoacrylate fuming.

While courts have widely acknowledged that photography is the most accurate way of documenting evidence for more than 100 years (Singh, 2012), the issue of accuracy (reliability) and bias (fallibility) has been heightened in the courts.


Figure 1—Take a look at squares A and B in the checkerboard image created by Edward H Adelson, a professor of vision science at MIT. They look like completely different shades of gray, right? Well, they are not. They are exactly the same color and shade!

“When judgments are made under uncertainty, two general types of errors are possible—false positives and false negatives. A decision maker cannot simultaneously minimize both errors because decreasing the likelihood of one error necessarily increases the likelihood of the other.” (Green, 1989)

There are a number of photographic processes that can also be used to manipulate an image “creatively”. For example, a photograph can be manipulated by changing exposure, camera angle, or choosing the wrong (improper) lens. Someone viewing a picture of a small hole in the ground could be easily misled by using a wide-angle lens, thus making the hole appear as a large abyss.

Fortunately, digital cameras provide data that can help identify whether an original image has been altered or if it is realistic. All digital cameras manufactured within the past decade capture information about the image pixels (also known as picture elements) together with information such as the date and time the image was taken; camera make, model, and serial number; and camera settings such as aperture, shutter speed, ISO, exposure compensation, lens used, and focus distance. These data elements, also known as exchangeable image file format (Exif) data, are stored as part of the image file in a collection of data fields called the file header, or metadata. Maintaining this data has become a requirement for all evidentiary images, which also provides interoperability between digital cameras and image processing programs, such as Adobe Photoshop. Exif data has also proven to be very useful for accurate analysis and reaching an accurate conclusion.


Figure 2—Just like in the Adelson image, ridge detail appears to be a different color because of the background contrast, which can create confusion when trying to identify ridge events.

In addition, the imaging sensors in most consumer-grade digital cameras today can detect in excess of 265 separate, distinct shades of gray. The imaging sensor in most professional digital single-lens reflex (DSLR) cameras can distinguish between 4,096 individual gradients for 12-bit dynamic range or 16,384 individual gradients for 14-bit dynamic range.

Digital images consist of pixels where each pixel has a specific color value, such as teal, fuchsia, orange, or brown, based on the hue, saturation, and brightness values of red, green, and blue for each pixel. Most digital images are stored as 24-bit color (8 bits per channel), which provides 256 possible shades of red, 256 possible shades of green, and 256 possible shades of blue for a total of 16,777,216 possible color values for each pixel in the image.

Some people argue that capturing anything other than 256 shades per color channel is a waste of file space because the human eye cannot see that many different shades, and many video cards and monitors do not support the display of that many different color values. It should be remembered, however, that applications like Adobe Photoshop have the ability to process up to 65,536 different shades per channel, which can be crucial when trying to suppress background noise.

Computer screens also have a pixel-based resolution, but that resolution has no direct correlation to the pixels in a digital image. For example, typical monitor resolutions today range from 1024 x 768 pixels to 1920 x 1200 pixels, and can be 17, 24, 27, or 29 inches wide (diagonally).

The monitor resolution refers to the number of color pixels that can be represented on the screen, which are also made up using a combination of red, green, and blue light values. Although meaning the same thing, these light values (a.k.a. picture elements) are displayed using different technologies such as liquid crystal display (LCD) technologies, light-emitting diode (LED) technologies, or gas-plasma technologies.

The end result is that a combination of red, green, and blue lights coupled with an intensity or brightness value of each individual light creates a display that causes the human eye to perceive a specific color value for the pixels that comprise the digital image.

This perception is also based on a “resampling” of the actual image data when the image is viewed on the screen. The resolution of digital cameras used by law enforcement agencies vary greatly:

  • 6016 x 4016 pixels (24 MP)
  • 4928 x 3280 pixels (16 MP)
  • 3968 x 2976 pixels (12 MP)
  • 3872 x 2592 pixels (10 MP)
  • 3072 x 2048 pixels (6 MP)

While considering the resolution of the image, the resolution of the monitor (as well as the video card or video driver) must also be taken into consideration. For example, a monitor can be set to a resolution of 1920 x 1200 pixels, 1600 x 1024 pixels, or 1024 x 768 pixels.

In other words, resolution by itself does not imply image quality or output size because the same number of pixels can be displayed as a small area (when zoomed out) or as a large area (when zoomed in) on a monitor or printed image. For example, minute detail can be overlooked when 6016 pixels horizontally are resampled to approximately 1600 pixels horizontally, which means that one pixel displayed on the monitor represents almost four pixel values averaged together. When images are resampled for printing, the loss of image quality and detail is far greater due to the technology limitations of today’s printers.


Figure 3—The analysis of an image (input) must account for biases and other factors (internal state) that can influence the response/conclusion. This includes the proper interpretation of data values based on experience, skill and training as well as other external issues/pressures, such as the type of crime, etc. when reaching a conclusion.

Historically, forensic experts—such as latent print examiners, footwear and tire tread examiners, and questioned document examiners—have relied on one-to-one prints for comparison. Regardless of the type or model of printer used today, the image quality is not as good as the traditional film-based photographs used in the past. With traditional photographic prints, there are no dots or pixels that have to be dithered for a clear, precise image, especially if the image is printed as a 1:1 life-size image. Today however, if a latent print is scanned on a flatbed scanner with a resolution of 1200 PPI, and the output device prints 500 PPI, then the latent print image must be resampled to eliminate a total of 700 pixels. This means that more than 50% of the actual pixel information is lost when the image is printed.

It also means that latent print examiners must either retrain their eyes to use lower-quality images for comparison; to do comparisons on the screen; or to use images printed at 2, 3, or 5 times life size.

Many forensic experts have found that it is significantly easier to view the images displayed side-by-side on a computer screen where they also have the ability to zoom in and out of the image for a more accurate and reliable interpretation of the actual image data.

We cannot, however, forget that the accuracy of every analysis and conclusion of any image begins with accurate and reliable image capture; therefore, the goal is to capture the best possible image with the highest possible resolution.

To capture an object with a digital camera with the highest possible resolution, fill the frame with the object and a scale. Failure to optimize image resolution creates problems when attempting to suppress background patterns, such as backgrounds on checks or money orders. As the size of the area of capture increases, the camera must be moved further back from the object, and a single photoreceptor must capture a larger area, using all the color values in the area covered by that single sensor and the light intensity to determine the resulting pixel value recorded by that photoreceptor.

One major consequence of trying to capture an area that is too large with a digital camera is that the optics in the camera blend multiple color values (from multiple source elements) together as a single pixel. In addition to losing detail and sharpness in the image, other problems are created, such as moiré patterns. By capturing a smaller area, the area covered by each photoreceptor is minimized, which provides clearer (and sharper) image detail, which also provides better image quality because the more accurate color values provide greater image detail.

Even with all the sophisticated tools and features in Adobe Photoshop, the image quality of a low-resolution image cannot be improved. That only happens on CSI... and that is television. The most anyone can hope to accomplish with a low-resolution image is to make it look less ugly… Kind of like putting lipstick on a pig!

The bottom line is that with the proper digital image processing standards and guidelines in place, and with appropriate training (for the photographer, the person enhancing the images, and the person analyzing the images), there should be no questions about whether or not the original image was altered, or that the interpretation of the image data provided an accurate, reliable conclusion.


About the Author

David “Ski” Witzke is the vice president of program management for Foray Technologies. He has more than 20 years of AFIS and forensic digital imaging experience and has conducted hundreds of digital imaging training programs for law enforcement agencies throughout the U.S. and Canada.


References

Singh, H., P. Kumar, R. Nanra, A. Kumar. “Why Is The Crime Scene Photographed? … There Is Not A Single Answer!” The Internet Journal of Forensic Science (2012, Volume 5 Number 1). Retrieved from: http://ispub.com/IJFS/5/1/13735

Green, D.M., J.A. Swets. Signal Detection Theory and Psychophysics. Los Altos, CA: Peninsula Publishing (1989).

 
< Prev   Next >






Digital-Image Management at Mass Gravesites

SKELETONIZED REMAINS that were carefully unearthed from the desert sands of Iraq tell their own story: the bones of an adult, still dressed in a woman’s apparel, lie supine. The skull is perforated by a bullet hole. Tucked in the space between the ribs and the left humerus is a much smaller skeleton, bones in the skull un-fused, and the fully clothed body partially swaddled in a blanket.

Read more...