Bias and Uncertainty W hen judgm ents are m ade under uncertainty, - - PowerPoint PPT Presentation

Presented by: David “Ski” W itzke Vice President, Program Managem ent FORAY Technologies ski@foray.com w w w .foray.com Presented by: David “Ski” W itzke Vice President, Program Managem ent FORAY Technologies ski@foray.com w w w .foray.com

I nternational Sym posium on Forensic Science Error Managem ent Hosted by the National I nstitute of Standards and Technology ( NI ST) W ashington, DC July 2 1 – 2 4 , 2 0 1 5 I nternational Sym posium on Forensic Science Error Managem ent Hosted by the National I nstitute of Standards and Technology ( NI ST) W ashington, DC July 2 1 – 2 4 , 2 0 1 5

ERROR M ANAGEMENT I N FORENSI C D I GI TAL I MAGI NG …

A RESOLUTI ON FOR THE ENTI RE CRI MI NAL JUSTI CE COMMUNI TY

ERROR M ANAGEMENT I N FORENSI C D I GI TAL I MAGI NG …

A RESOLUTI ON FOR THE ENTI RE CRI MI NAL JUSTI CE COMMUNI TY

July 2 2 , 2 0 1 5 2

Bias and Uncertainty …

W hen judgm ents are m ade under uncertainty, tw o general types

• f

errors are possible – false positives and false negatives. A decision m aker cannot sim ultaneously m inim ize both errors because decreasing the likelihood

• f
• ne

error necessarily increases the likelihood of the other.

( David M. Green & John A. Sw ets, Signal Detection Theory and Psychophysics, 1 9 6 6 )

July 2 2 , 2 0 1 5 3

Bias and Uncertainty …

( From Macm illan & Creelm an, Detection Theory, 2 0 0 5 )

July 2 2 , 2 0 1 5 4

Are YOU biased?

( W hat do YOU see?)

Courtesy of w w w .1 2 3 opticalillusions.com

July 2 2 , 2 0 1 5 5

First, know your lim its and the lim its of technology…

July 2 2 , 2 0 1 5 6

Are blocks A & B the sam e shade of gray?

July 2 2 , 2 0 1 5 7

State of Connecticut versus Alfred Sw inton The Connecticut State Suprem e Court conclude that the trial court properly adm itted into evidence the com puter-enhanced photographs But im properly adm itted the overlays created in Adobe Photoshop because the expert relied on the overlays for his opinion but yet he had “no m ore than an elem entary fam iliarity” w ith the program used to create the overlays

I fought the law …

July 2 2 , 2 0 1 5 8

provide guidance for the acceptance

• f

forensic evidence based upon com pliance w ith standards:

( 1 )

The equipm ent is accepted as “standard” in the field

( 2 )

Proper procedures w ere follow ed

( 3 )

Reliable ( standard) softw are program s are used

( 4 )

The equipm ent is in good w orking order, and

( 5 )

The exhibit is properly identified as the output in question.

( 6 )

Operated by trained and com petent users

Federal Rules of Evidence …

( aka “Sw inton Six”)

July 2 2 , 2 0 1 5 9

But w e m ust understand som e basic technologies and term inologies to develop reliable standards. The m ost com m only used but m isunderstood term s are: Resolution Pixels Per I nch ( PPI ) Dots Per I nch ( DPI ) Sam ples Per I nch ( SPI )

They’re only w ords …

July 2 2 , 2 0 1 5 1 0

To err is hum an …

Techniques for error m anagem ent include:

• Use appropriate equipm ent/ file form ats
• Use appropriate techniques
• Collect standardized data
• Develop standardized reporting
• Extend peer review
• Develop a culture of quality im provem ent

July 2 2 , 2 0 1 5 1 1

CAPTURE ISSUES

July 2 2 , 2 0 1 5 1 2

W hat issues m ust be m anaged w hen photographing an evidentiary

• bject ( piece of evidence) ?

W hat is the resolution required to capture m inute detail in an object, such as a fingerprint, footprint, tire tread, questioned docum ent, autopsy photo, blood spatter, or other im age used for analysis, com parison, and conclusion?

• W hat is the resolution of the cam era or other capture device?
• W hat is the size of the area/ object being captured?
• W hat is the distance from the cam era lens to the object being

captured?

• W hat is the lens that is being used?

July 2 2 , 2 0 1 5 1 3

NOTE: I n photography, the inverse-square law m eans that the resolution is inversely proportional to the square of the distance from the source of that physical field of view .

from the lens to the object m ay vary, w hich m eans the resolution ( PPI ) w ill also vary significantly

The distance …

July 2 2 , 2 0 1 5 1 4

Validating your com puter and softw are settings …

Of your m onitor:

• 1 9 2 0 x 1 2 0 0 pixels
• 1 6 8 0 x 1 0 5 0 pixels
• 1 4 4 0 x 9 0 0 pixels
• 1 2 8 0 x 8 0 0 pixels
• 1 0 2 4 x 7 6 8 pixels

Of your im age:

• 6 0 1 6 x 4 0 1 6 pixels ( 2 4 MP)
• 4 9 2 8 x 3 2 8 0 pixles ( 1 6 MP)
• 3 9 6 8 x 2 9 7 6 pixels ( 1 2 MP)
• 3 8 7 2 x 2 5 9 2 pixels ( 1 0 MP)
• 3 0 7 2 x 2 0 4 8 pixels ( 6 MP)

W hat are the potential issues that you m ust consider w hen validating your cam era?

• W hat is the resolution of:

July 2 2 , 2 0 1 5 1 5

W hat are the potential issues that you m ust consider w hen validating your cam era?

• W hat is the size of the im aging sensor:
• 1 8 .7 x 1 4 m m

( 1 .8 x focal length m ultiplier)

• 2 2 .3 x 1 4 .9 m m

( 1 .6 x focal length m ultiplier)

• 2 3 .6 x 1 5 .7 m m

( 1 .5 x focal length m ultiplier)

• 3 6 x 2 4 m m

( 1 x focal length m ultiplier)

• And w hat effect does this have on your lens?

Validating your com puter and softw are settings …

July 2 2 , 2 0 1 5 1 6

To scan or not to scan … that is the question

July 2 2 , 2 0 1 5 1 7

All scanners are NOT created equal!

July 2 2 , 2 0 1 5 1 8

HARDWARE ISSUES

July 2 2 , 2 0 1 5 1 9

Validating your hardw are and softw are settings …

W hat are the potential issues that you m ust consider w hen validating your com puter:

• I s

your com puter equipped w ith an appropriate video card

• W hen

setting Preferences, w hich

• ption

provides the highest im age quality w hen displayed:

• Graphics Processor function Enabled
• Graphics Processor function Disabled

July 2 2 , 2 0 1 5 2 0

Validating your com puter and softw are settings …

Use a standardized test target in w hich you know precisely w hat level of detail is contained w ithin the im age file

July 2 2 , 2 0 1 5 2 1

Validating your com puter and softw are settings …

Edit > Preferences > Perform ance

• Graphics Processor Settings
• Use Graphics Processor function Enabled

July 2 2 , 2 0 1 5 2 2

Validating your com puter and softw are settings …

Edit > Preferences > Perform ance

• Graphics Processor Settings
• Use Graphics Processor function Disabled

July 2 2 , 2 0 1 5 2 3

F0RMAT ISSUES

July 2 2 , 2 0 1 5 2 4

RAW : RAW files are sim ply the unprocessed sensor data from a cam era

12 – bit Grayscale = 4,096 shades of gray 14 – bit Grayscale = 16,384 shades of gray 16 – bit Grayscale = 65,536 shades of gray

JPG: An im age form at that reduces the im age file size – sacrificing im age quality – so that the im age does not require as m uch storage space

8 – bit Grayscale = 256 shades of gray

TI F: A non-com pressed form at used for storing digital im ages

8 – bit Grayscale = 256 shades of gray

Common camera formats …

July 2 2 , 2 0 1 5 2 5

Com pression – a process for reducing the size of a file to facilitate the storage and transfer of large files. Lossless Com pression – a form at ( such as RAW

• r TI F)

used for reducing the size of a file w here no inform ation is lost, but the com pressed file m ay use few er bits to represent the inform ation. W hen the file is re-opened, the

• riginal data is reconstructed.

Lossy Com pression – a form at ( such as JPG) used for reducing the size of a file w here inform ation is lost and cannot be retrieved in its original form .

Using the appropriate file types

July 2 2 , 2 0 1 5 2 6

Altered or Com pressed?

Digital im ages should not be com pressed to save space unless a loss-free ( aka lossless) m ethod is used – lossy ( aka lousy) com pression is NOT an acceptable m ethod! NOTE: JPG com pression rem oves pixel values, relocates pixel boundaries, alters pixel colors, and reduces resolution differently for different locations in the sam e im age.

July 2 2 , 2 0 1 5 2 7

PDF Com pression

W ARNI NG: Using the standard default PDF settings, im ages w ith a resolution greater than 2 2 5 PPI are dow n-sam pled to

• nly 1 5 0 PPI , and the dow n-sam pled im ages are then

com pressed using standard JPG com pression w ith only m edium im age quality.

July 2 2 , 2 0 1 5 2 8

PDF Com pression

July 2 2 , 2 0 1 5 2 9

PDF Com pression

July 2 2 , 2 0 1 5 3 0

PROCESSING ISSUES

July 2 2 , 2 0 1 5 3 1

Use Non-destructive Techniques …

Use im age processing techniques that are non destructive, such as:

• Black & W hite
• Curves
• Shadow s/ Highlights
• Cam era RAW Filter

July 2 2 , 2 0 1 5 3 2

Use I m age Processing Techniques Properly …

July 2 2 , 2 0 1 5 3 3

Don’t get frustrated … or rotated!

Rotating an im age at any degree other than 9 0 degree increm ents causes a num ber of artifacts that can affect digital im age processing.

July 2 2 , 2 0 1 5 3 4

Verify your com puter and softw are settings…

I m age > I m age Size > Resam ple

• 5 0 0 PPI to 3 0 0 PPI
• Autom atic
• Preserve Details
• Bicubic Sm oother
• Bicubic Sharper
• Bicubic
• Nearest Neighbor
• Bilinear

July 2 2 , 2 0 1 5 3 5

Verify your com puter and softw are settings…

July 2 2 , 2 0 1 5 3 6

Verify your com puter and softw are settings…

July 2 2 , 2 0 1 5 3 7

OUTPUT ISSUES

July 2 2 , 2 0 1 5 3 8

Validating your com puter and softw are settings …

W hat potential issues that you m ust consider w hen validating your output devices:

• W hat output device are you using:
• Dye sublim ation printer
• I nk jet Printer
• Laser Printer

July 2 2 , 2 0 1 5 3 9

A true and accurate representation?

July 2 2 , 2 0 1 5 4 0

Validating your com puter and softw are settings …

W hat potential issues that you m ust consider w hen validating your output devices:

• W hat are the specifications of your output device:
• Num ber of dots, nozzles, droplets
• Size of droplets ( picoliters)
• Num ber of I nks
• Type of paper used

July 2 2 , 2 0 1 5 4 1

Device dot clusters

July 2 2 , 2 0 1 5 4 2

The num ber of dots …

change w hat w e “see” in the printed im age!

July 2 2 , 2 0 1 5 4 3

Thank you very m uch for your attention and for your interest!!!