Calidad

Software for automated imageto-image co-registration
Cody A. Benkelman, PI
Heidi Hughes, President
Spatial Technologies, LLC
Cape Canaveral, Florida

Poster presentation created for
The Civil Commercial Imagery Evaluation Workshop
March 14-16, 2006

Software for automated imageto-image co-registration
This presentation represents an interim status report.
Our SBIR Phase II project isin its second year, and is
scheduled for completion at the end of 2006.
Supported by: NASA SBIR Program
Phase I: 2003
Phase II: 2004 – 2006
Contract #NNS05AA07C
NASA Stennis Space Center
COTR: Bill Graham

Basic Premise
In spite of excellent quality in state of the art
orthorectification software, camera models, and digital
elevation models, two orthorectified images of the samearea often exhibit minor errors relative to each other.
Although the absolute geometric accuracy is typically
excellent, the relative accuracy of the image-to-image coregistration can be improved. For automated change
detection at the scale of one pixel, the relative accuracy of
two independently created orthophotos is often inadequate.
The fundamental goal of this SBIR Project is to developsoftware that applies multiple techniques to achieve
subpixel precision in the co-registration of image pairs.

Project Objectives
• Develop software to fine-tune image-toimage co-registration, presuming images are
orthorectified prior to input.
• Create a reusable software development kit
(SDK) to enable incorporation of these tools
into other software.
• Provide automated testing forquantitative
analysis.

Project Deliverables
• Software development kit (SDK),
developed in C++ (ANSI standard)
• Windows application
– For automated testing (.NET framework)
– Example application for SDK implementation

Features & Benefits
of image co-registration tools
FEATURES
• Will provide for very accurate (sub-pixel)
image-to-image co-registration, assuming
input image data isalready well
registered.

BENEFITS
• Improves the accuracy of automated change
detection.

•

SDK will allow incorporation of these
tools into other software.

•

Enables a User with a need for these tools to
customize them for their application.

•

Test system (Windows .NET compliant)
will provide ability to compare
effectiveness of alternative methods.

•

Enables userto add new technology and
determine if it improves the results in their
application.

Software Architecture

Primary Functions – SDK &
Windows Application
• Change Detection
• Identification of tie points
– (a.k.a. matching points, landmarks, pass points)

• Automatic measurement of initial and final
accuracy
• QC tools

Basic Software Flowchart
Begin
coregistration
processImage A

Image B

Configure software
and read image pair

Identify areas of
significant change

Apply co-registration
accuracy test

Accuracy
"BEFORE"

Co-registration
accuracy improved?

Conduct tie point tests,
select passing points

Yes

Rectify Image A
to match B
Calculate local
warp coefficients

Repeat co-registration
accuracy test

No

Accuracy
"AFTER"TBD!

Process
Complete

Image Pair QC tools

Side by side tie point view

Image Pair QC tools

Residual tie point error view

Image Pair QC tools

Overlay view

Reporting Functions

Added Functions – Development
& Validation Tools
• Recipe Builder
– For algorithm development
– Intuitive GUI for linking core processes
together into complex algorithms
– User canquery & modify process data

• Recipe Validation
– Batch processing of recipes against multiple
input datasets, for statistical analysis

“Recipe Builder”
Each process can be
queried (input & output
data, configuration
data) for testing, and
data may be modified
to alter the results of
subsequent processes.

Note the “look and feel”
of this UI are preliminary…

Recipe Validation...