Area Ratio Coregistration/Conflation (ARC) tool

Welcome to the download page for ARC Plugin for ESRI ArcMap (Currently Version 1.5).

The Plug-in adds a toolbar into the ArcMap environment which allows the user to conflate or co-register two raster images loaded as raster layers with a single click of the conflate button. The co-registered/conflated image is inserted as a third layer and shown in a transparent overlap with the image/raster layer to which it was co-registered.

The ARC Plug-in does not need any estimates of unknown parameters of scalings, rotations, shifts and skewings or images to run.

Images should have a common part that includes at least three closed regions in common that are not at the edge of each image.

Currenly the co-registration process is automatic for limited types of raster images. While the process is automatic only limited types of raster images are currently automatically coregisterable.

The supported image types include:
1) arbitrary "cuts of the same image" that can be differently resized (zoomed), rotated, shifted, and skewed (see example 1). The ARC Plug-in was extensively tested with on images, which are different cuts of the same image cuts that are differently affinely transformed. Truly automated co-registration/conflation succeeds reliably in these cases.

2) Images that are not cuts of the same image, but are well segmented by color (e.g., sharp maps with few colors and large unicolor regions)(see example 2).
Limitation: Extra small regions, or minor noise on regions that are standing out by their size, do not cause the method to fail, although large numbers or similarly sized regions and major changes or noise will. Minor changes on similarly sized regions can also cause failure.

3) Vector images (see example 3) can also be conflated by taking a screenshot of or by rasterizing the vector data by some other method to produce a BMP file to be used as input for the ARC Plugin. The requirements of having 3 closed regions that are not at the edge in common between the images, as well as not having large numbers or similarly sized regions or major changes and noise, still apply to rasterized vector data as they do to the two raster types above. Fortunately, by definition, rasterized vector data has less of these problems than an unrestricted raster. Future versions of the Plug-in will work with vector images (ESRI shape files) directly without rasterizing.

Invariance to affine transform

Area Ratio Algorithm is invariant to a general affine transform including disproportional scaling. The algorithm was extensively tested on images which are different cuts of the same image including affinely transformed cuts. Truly automated conflation succeeds reliably in these cases. Minimal feature richness of images is still required for "cuts of the same image" cases. Three (3) closed regions in the overlapping areas of the 2 images are required. Regions that touch the edge of the image are not counted as closed.

The Area Ratio Algorithm is a formalization of human manual conflation(co-registration) process as practiced by a human. Only a few of a sizable number of steps in the human process have been formalized and are part of the algorithm. Nevertheless these few formalized steps already allow the automated conflation of the above image types. As a result of this approach, all actions of the algorithm can be interpreted in terms of notions from the human process. The ARC algorithm mimics the human co-regstration process. This means that if a human cannot reliably conflate/co-register images, the ARC algorithm may also not be able to do this reliably.

Short usage sequence

Two raster images are added as raster layers into an ArcMap document. Initial coordinates for both images are set up formally to match top left corners. This typically is not a correct match. Both rasters initially have world files with zero coordinates for top left corners. After the images are loaded into Arcmap, a single button click on the toolbar runs the ARC algorithm which automatically conflates the images and produces a world coordinate file which contains the relative transform, and the second image is also warped to match the first one based on the found relative transformation.

Longer detailed usage sequence (from online help)

If you would like a screenshot illustrated procedure or are having a problem, go to the Illustrated Usage Sequence where every step in the sequence has an associated screenshot.

Steps for running ARC plugin for ESRI ArcMap:

1) Run ArcMap

2) Start with a new empty map

3) Click Layers->Add Data

4) Browse to and select the first bmp file (*.bmp) (for example image1.bmp)

5) Browse to and select the second bmp file (for example image2.bmp)

Note: A user creates image1.bpw and image2.bpw files. In ArcMap 9.0 image1.bpw and image2.bpw files need to be in the same directory as image1.bmp and image2.bmp. Both image1.bpw and image2.bpw need to contain the y-flipped identity matrix.

The following are the needed contents of bpw files (y-flipped identity matrix):
1.0
0.0
0.0
-1.0
0.0
0.0

The y-flip is needed to correctly display bmp files in ARCMap, when the coordinate system origin for Windows bmp files is in the top left corner with positive y-axis going down, and the coordinate system origin for ArcMap is in the bottom left corner with positive y-axis going up.

6) Right-Click the first created raster layer (named Image1.bmp in the example) and select Properties in the menu.

7) On the Layer Properties dialog box that appears select the Display Tab, and in the Transparent field change 0% to 50%. Click OK.

8) Change the transparency of the second raster layer to 50% as well using the same click sequence.

9) Press the Conflate button "C" while leaving Visualization "V" button in the pressed position to run the algorithm.

10) Visualization will play, and, if an accurate match is found, the matching shape pairs message box showing which shapes were chosen and then centerpoint coordinates message box will appear showing coordinates (Centers will also be marked by red and blue squares on the main visualization window).

11)Click OK on both message boxes. The visualization window will disappear and the result files are created. The results files are Conflated_image2.bmp and Conflated_image2.bpw, where image2.bmp is the filename of the second image that was added to the map document. Conflated_image2.bpw file contains the relative world coordinates found. The transformed Conflated_image2.bmp is added as a 3rd layer to the map. It is shown transformed based on the found relative transform as well as 50% transparent, which allows one to see the overlapped conflation result.

Supported Image formats

Only BMP(bitmap) and JPG(JPEG) with associated BPW and JGW world files are currently supported as raster layers. BMP is the preferred type because JPEG compression adds an extra layer of smoothing/distortion to the image.

New in version 1.5. Version 1.5 now includes support for TIF and associated TFW files as raster layers.

Install/Uninstall procedure

To install, unzip ARC_Conflation_Plugin_for_ArcMap_09Demo.zip into a new directory. Run Setup.exe. Follow onscreen instructions. Once the installation is finished the shortcuts to help will be in the Start Menu and the Plug-in toolbar will be available in ArcMap. To Uninstall go into Start->Settings->Control Panel->Add/Remove Programs. Find Conflation Plugin in the list of programs and remove it.

Demo Version

The demo version limits the size of the input images to images no larger than 300x300 pixels. Other than the size limit the demo is fully functional.

Click the following link to download the demo version 0.9.

Full Version

Send an email to Imaglab@cwu.edu with your contact information if you are interested in the full version without the demo limit.

Samples Collection

The collection of sample images which the ARC Plugin is able to automatically co-register/conflate is available as a collection of HTML pages. The collection includes screenshots taken before and after co-registration/conflation.

Click the following link to download the ARC Plugin Samples Collection.

Feedback

We would appreciate your feedback sent to Imaglab@cwu.edu. We also would be interested in analyzing images that were or were not conflated/co-registered properly.

New Version 1.5

Version 1.5 introduces Cross-Line Length Ratios in addition to the Area Ratio as an Invariant image analysis parameter. This addition enables successful conflation in cases with large numbers of similarly sized shapes or objects. With this addition, the algorithm takes into account the form or shape of an object not just its area. This improvement requires extra cross-line length data collection, and therefore the algorithm takes a bit more time than the version 0.9 for the cases with distinctly sized shapes that version 0.9 is able to conflate correctly.

Version 1.5 introduces the option of manual contouring/pre-segmentation as opposed to the default automatic contouring. The color to be used as the manual contour color is specified in the Settings dialog. The regions outlined manually with the specified color are used directly in the analysis. (A more detailed description of manual countouring is in the Settings).

Version 1.5 also introduces support for TIF images and associated TFW files to be used as input raster layers.

Click the following link to download the demo version 1.5.