DPT Central Lab Photofinishing Field Training

1
DPT Central Lab Photofinishing Field Training

• Image Processing Overview Image Hub and Photo
  Imaging Engine
• Jon Lewis, RD Project Manager

2
Topics Covered

• Key terminology
• Overview of Image Hub and Photo Imaging Engine
  components
• Network connectivity
• The rack
• As the image passes through the system
• Aspect ratio CNF vs. DSC
• Laying out mixed orders
• Image Quality
• 2003 Plan of Record
• 2004 and Beyond
• Beyond MAT1 PfDS
• Prints for Portrait Labs
• Prints for Reprints Enlargements

3
Key Terminology

• Photofinishing Terms
• PfDF Photofinishing Data Format (Industry
  standard)
• A batch is N number of customer orders.
• Batch information described in PFDF-like schema
• Hierarchical
• Text based
• A customer order is 1..M output sizes of 1..X
  images.
• BMC/BMS/LDM Batch Management Computer/System,
  Lab Data Management
• Computing infrastructure that controls the
  factory floor, and provides PfDF descriptions of
  work
• Typically done using data bases, and are fully
  redundant
• Our Indigo Terms
• A job is a JLYT job which is defined by the SW
  system.
• An element is a single image (print, I-mark )on
  a spread.
• A spread is defined by a physical page.

4
Image Hub Overview

• Standard off-the-shelf dual processor Intel
  workstation running Windows XP.
• Provides the following key features/functionality
• Connectivity into a customers lab data management
  system for the management of batches/jobs to be
  processed.
• Connectivity into a customer image data server to
  retrieve images (pictures) to be processed.
• Coordinates the delivery of image data to one or
  more Photo Imaging Engines (PIEs) for
  processing.
• Optimizes the layout of customer prints on 12
  wide web based upon customer order information.
• Converts image data into a common LAB or RGB
  format from a variety of input formats (TIFF,
  JPG, ).
• User interface provides basic status of press,
  error recovery functions, user customization,
  startup/shutdown.

5
Photo Imaging Engine Overview

• Standard off-the-shelf dual processor Intel
  workstation running Windows XP. Default
  configuration of solution will include two PIE
  systems.
• Provides the following key features/functionality
• Final color space conversion to 6-inks (KCMYcm).
• Compress 6-ink data into Indigo Compression
  Format (ICF).
• Connection to c9100 press to deliver image data
  in ICF data format.
• Highly performance optimized image enhancement
  pipeline that includes sharpening/smoothing,
  scaling, rotation, cropping, and contrast
  enhancement.
• Connection to Image Hub to receive image data and
  to provide status.
• User interface control is provided through Image
  Hub.
• Treats a single image as a unit of work. Can work
  on multiple images at a time due to its use of
  multi-threading.
• Best to think of the PIE as an photo image RIP.

6
Physical (Network) Connectivity
Gigabit
Megabit
7
The Rack

• Image Hub and PIE computing systems will mounted
  in a Compaq Series 10000 rack
• Rack will include Image Hub and PIE computers,
  network switches, 15 display, keyboard, and
  pointing device.
• Has capability to house 3 additional computing
  systems
• Scalability for future Photo Imaging Engines.
• If we need more than additional 3, we can go to a
  bigger rack.

8
As the image passes through the system
9
Aspect Ratio CNF vs DS

• CNF is 32 aspect ratio
• For 4 print, 4x6 is great, simply double in both
  directions
• DSC is 43 aspect ration
• DSC for 4 print, 4x5.33 is ideal no data loss
• To get a 4x6 print, you would need to crop ¼ inch
  from top and bottom
• For MAT1, we will support no loss DSC sizes (see
  MAT1 data sheet)

Preserve 6
Preserve 4
10
Basic Pipeline
Image Enhance Render Color Match Compress
Hardware Based
Screening
Software Based
INK on PAPER
11
What Does an LEP Pixel experience before the
writing head burns it?
conversion into Lab color space
Image Enhancement Filters
Color Matching Table
12
Color Matching Table

• What it is
• 3D trilinear Interpolation used to calculate
  values between nodes
• Color Matching Table is a 256 cubed space that
  represents Lab to CMYKlclm
• How it is created
• Theoretical profile (gamut map) of sRGB to CMYKcm
• Neutrals and near neutrals calibrated to Delta E
  specifications
• ICC Profile (Lab to CMYKcm) for perceptual
  rendering intent
• Tone mapping place to fine tune skin tones
• Color conversion color map ? color matching
  table

13
Mixed Order Layout

• Complexity of the system / key part of value
  proposition
• Factors that affect the algorithm
• waste minimize or who cares
• layout style template vs. dynamic
• input image constraints any size vs. fixed sizes
• print layout order small-to-large,
  largest-first...
• repeat length variable vs. fixed
• printable width
• landing zone yay or nay
• eye mark yay or nay
• calibration jobs yay or nay
• performance fast or slow

Sample order 1 8x10 4 5x7 12 4x6
14
In Summary for a given image

• Image Hub
• Gets image file name from PfDF batch description
• Pulls image file from image server
• Convert image to private color space (LAB)
• Determine final layout location for image
• Pass image data to Photo Imaging Engine
• Photo Imaging Engine
• Pass image data through image pipeline
• Scale/rotate
• Sharpen/smooth
• Contrast enhance
• Color match/compress
• Pass compressed image data to press

15
Imaging for 2003
16
Sharpening

• Selectively sharpens edges and fine detail.
• Avoids sharpening noise.
• Brings out feature detail by using a local area
  edge filter.

17
Smoothing

• Selectively smoothes noise due to
• Poor capture devices.
• Compression artifacts.
• Avoids smoothing edges or fine detail.

18
Automatic Contrast Enhancement

• Brings Depth to faded or foggy images
• Fixes under/over exposure.
• Improves contrast.
• Does not alter hues.

19
UI Controls for Color/Imaging

• Slider controls for
• color balance
• Individual controls for W, R, G and B to control
  tone balance (e.g. allows customer to generate
  warmer/cooler colors in output)
• Contrast adjustment
• Sharpness adjustments
• Enable/disable switch for each adjustment
• Others(?)
• Will be exploring other required controls with
  CeWe in the coming months.

20
LEP vs AgX Color Gamut

• LEP coverage of AgX
• Fuji Crystal Archive 94
• Kodak Duralite 92
• AgX coverage of LEP
• Fuji Crystal Archive 76
• Kodak Duralite 87
• In theory
• AgX better in some natural colors (flowers,
  skies) due to more blue/magenta gamut.
• LEP is better in Caucasian flesh tones due to
  more yellow/red/orange gamut.
• HOWEVER, we have not seen perceptual proof of
  this to-date.

LEP Color solid Silver Halide Wire mesh
21
LEP vs AgX Its in the Print

• Permanance factors
• Laminated output will perform better that
  traditional AgX (e.g. you can simply wipe things
  off our laminated output)
• Lightfastness of print is very competitive and
  early results show in some cases better than AgX.
• Early customer feedback based upon images printed
  at Photokina 2002
• The overall quality of the hp prints is very
  good. None of them would be rejected by a
  customer of regular printing services. Mike
  Stroud, Technical Director, Pixology Ltd.

22
Imaging for 2004 and beyond
23
Improved Scaling
Pixel Replication
With Smart Focus
24
Automatic Red-Eye Removal

• Automated
• treated as a black box within our enhancement
  pipeline
• Accurate
• obtains gt75 id / correction rate
• minimizes false positives
• Fast (relative to current DeskJet requirements)

25
Imaging Tuned for Portrait Customers

• Contrast enhancement
• Selective sharpening/smoothing
• Shadows/highlight management

26
Image Analysis

• Scene detection
• Identify types of areas within a scene (e.g.
  face, background).
• With this information
• Can selectively run certain image enhancements
• More aggressive approach to improving customer
  image

27
Building upon MAT1 PfPort PfRE
28
Prints for Reprints Enlargements

• Areas of SW Development
• Research available reorder stations (default
  Spitzke Fall 03)
• Optimize image processing algorithms for scanned
  film
• High probability that this will be a new scanner
  that we have not dealt with yet
• Understand new layout requirements
• Backprinting
• Integration to customer BMC/LDM

29
Prints for Portrait Labs

• Areas of SW Development
• New input formats
• New scanned film path
• High resolution digital source path
• Integration into LDM (Kodak DP2)
• New color management requirements
• Portrait labs have much tighter control, and
  place much higher emphasis on skin tones

30
(No Transcript)