Difference between revisions of "Registration: Manual 2D Series"
m (→Special notes regarding the number of references points) |
m (→Special notes regarding the number of references points) |
||
Line 36: | Line 36: | ||
== Special notes regarding the number of references points == | == Special notes regarding the number of references points == | ||
− | + | The following notes apply: | |
*For 2D images, three or more points are required. The algorithm may fail if nearly all of the points fall on the same line. | *For 2D images, three or more points are required. The algorithm may fail if nearly all of the points fall on the same line. |
Revision as of 17:05, 11 July 2012
The Manual 2D Series algorithm allows you to manually or semimanually register two 2D grayscale or color images by placing corresponding points on both images and then applying either the Landmark-Thin Plate Spline or Landmark-Least Squares algorithm. After applying one of these algorithms, you can then, if needed, use the other algorithm. In addition, at any time, you can manually adjust the alignment of each image.
The algorithm also lets you manually register different slices of a 2.5 image dataset in the same way as the 2D images.
Contents
- 1 Background
- 2 Image types
- 3 Special notes regarding the number of references points
- 4 Applying the Manual 2D Series algorithm on 2D images
- 5 References
Background
Through the Registration: Manual 2D Series window (Figure 1), this algorithm provides the tools for manually registering images and the ability to use two of the landmark registration algorithms. The Registration: Manual 2D Series window includes two tabs: Blended and Dual.
Blended tab
On the Blended page, the window superimposes the image to be registered on top of the reference image. This page includes icons for aligning, rotating, alphablending, and magnifying or minifying the images as well as icons for modifying the lookup table for each image and for applying a checkerboard pattern to assist in alignment.
Dual tab
Clicking the Dual tab opens the Dual page, which displays the images side by side. A number of icons appear in the toolbar. The red icons are color coded to the reference image, which appears on the left and is surrounded by a red border. The green icons apply to the image to be registered, which is shown on the right side of the page and is surrounded by a green border.
Delineating corresponding points
Before being able to use either the Thin Plate Spline or Least Squares algorithm, you must first delineate a minimum of three corresponding points on both images. These points serve as guides, or landmarks, for aligning the images. In order for the points to be corresponding, you should delineate each point in the same sequence; i.e., the first point that you delineate on the reference image should be the first point delineated on the image to be registered, the second point on the reference image should be the second point delineated on the image to be registered, and so on.
To delete a point, you must first select the point and then use the appropriate Delete icon to delete it. However, before you can select the point, you need to return the cursor to its default mode using the icon.
On both the Blended and Dual pages, you can modify the lookup table (LUT) by clicking , the LUT icon. When you select , the Lookup Table window appears. This window includes two tabbed pages: one for Image A, the reference image, and one for Image B, the image to be registered.
A common icon- , the Reset icon-appears on both the Blended and Dual pages. Use this icon if you want to restore the original version of the images. This icon deletes all points on the reference image and on the image to be registered.
Image types
You can apply this algorithm to 2D grayscale and color images.
The dimensions or image type of the image to be registered need not be the same as the dimensions or image type of the reference image. The registered image has the same image type as the image to be registered and the same extents as the reference image.
Special notes regarding the number of references points
The following notes apply:
- For 2D images, three or more points are required. The algorithm may fail if nearly all of the points fall on the same line.
- For 3D images, four or more points are required and the algorithm may fail if the points nearly all fall on the same plane.
Applying the Manual 2D Series algorithm on 2D images
Note:
Making manual adjustments to the image to be registered
To check registration, do the following:
Note: The label "0.75A" means that 75 percent of the adjusted image appears and only 25 percent of the reference image is displayed.
Note: Actually, you can change the color of the image to any color. The purpose in doing this task is to easily distinguish the adjusted image from the reference image.
You may want to maximize the window and enlarge the images so that you can view details of the images while checking alignment. To maximize the window, simply click File:ButtonRestore.jpg , the Restore button. To maximize the image, click File:RegistrationManual2DSeries54.gif , the Magnify image 2.0x icon, which magnifies the image to twice its current size, as many times as needed.
Moving the adjusted image in any direction
To move the adjusted image, click , and, holding down the left mouse button, drag the icon in the direction in which to move the image, and then release the mouse button. After a few moments, the image appears in the new position. Note that, depending on where you moved the adjusted image, its outer portions may disappear beyond the outer edges of the reference image (see Figure 3).). In this case, you can use to move the adjusted image again.
|
Moving the adjusted image with the direction arrows
The icons that display direction arrows ( , the Move image up icon; , the Move image down icon; , the Move image right icon; and , the Move image left icon) let you move the adjusted image in one direction only by a specified number of pixels. The default number of pixels is 1; however, the number of pixels may be from 0.01 to 2048.0. Therefore, the direction arrows provide the greatest specificity in moving the image, which is perfect for fine adjustments.
To change the number of pixels, click , the Set pixel increment icon; enter a number in the Pixel Increment box; and click Apply.
Rotating the adjusted image
To select the center of rotation for the image, click and set the center by clicking with the mouse on the image. The letter C appears with a yellow crosshair cursor (see Figure 4]). You can then drag a line from the center of rotation to move the image above or below the center of rotation. To rotate the image clockwise, click ; and to rotate the image counterclockwise, click .
As with the direction arrow icons, you change the degree increment to achieve exact registration. The default degree increment is 1.0, but you can specify from 0.01 to 360 degrees. To change the degree increment, click , the Set degree increment icon. The Change Degree Increment dialog box appears. Enter the degree increment in the Degree Increment box, and click Apply.
Registration: Manual 2D Series: Blended tab
Manual 2D Series: Dual tab
Applying the Manual 2D Series algorithm on 2.5 images
|
|
References
For more information, refer to:
K.S. Arun, T.S. Huang, and S.D. Blostein, "Least-Squares Fitting of Two 3-D Point Sets" , IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-9, No. 5, September, 1987, pp. 698-700.
David Eberly, "Thin Plate Splines" of Magic Software at http://www.magic-software.com. Also at the same site are the files: MgcInterp2DThinPlateSpline.h, MgcInterp2DThinPlateSpline.cpp, MgcInterp3DThinPlateSpline.h, MgcInterp3DThinPlateSpline.cpp Paper that explains warping from one x,y source to another x,y target set.
Fred L. BookStein, "Principal Warps: Thin-Plate Splines and the Decompositions of Deformations," IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 11, No. 6, June 1989, pp. 567-585.