DE10251217B3 - Automatic calibration system for multi projector system e.g. for passive stereo images, using generation, recording and image filtering of strip patterns - Google Patents

Abstract

The automatic calibration system uses at least 2 projectors, a digital camera and a control unit linked to the digital camera and the projectors, for generation, recording and image filtering of strip patterns, which are projected onto the largest possible projection surface via the projectors, for evaluation of the warp errors and image warping.

Description

introduction

The extension concerns a procedure for the automatic calibration of a multi-projector system that allowed any number of images from different Projectors are generated, superimposed on each other to the pixel Total luminosity to increase or enable a stereo display.

Problem Description

Multi-projector systems for display of information on big projection are known. An example of this is described in WO-A-00/18139. The projection images of the projectors are arranged side by side.

The overall brightness of an image to increase or a passive stereo image to produce, it is necessary that several projectors use the same Area of the screen illuminate. The problem with this image display is that the individual images of the projection units are usually not one above the other lie. This calibration is currently done manually and is associated with great effort and too imprecise.

The object of the invention is a to implement pixel-perfect auto calibration, which is the cumbersome manual fine adjustment is unnecessary. This calibration also equalizes still by the geometry of the projection surface or by the projector (Keystone) image distortion caused by a camera.

• – mindestens zwei Projektoren,
• – eine Digital-Kamera und
• – einer Ansteuerungseinheit zur Ansteuerung der Projektoren und der Kamera.

To achieve the above object, the invention proposes a multiprojector autocalibration system which is provided with

• - at least two projectors,
• - a digital camera and
• - A control unit for controlling the projectors and the camera.

• – Erzeugung, Aufnahme und Bildfilterung von Streifenmustern,
• – Finden der größtmöglichen gemeinsamen Projektionsfläche und
• – Berechnung der Warp-Felder und Imagewarping.

The auto calibration is carried out according to the following steps:

• - generation, recording and image filtering of stripe patterns,
• - Find the largest possible common projection surface and
• - Calculation of warp fields and image warping.

The following are the sub-steps auto calibration are briefly explained using the illustrations.

In detail show:

1 Experimental setup with two projectors,

2 Background image (projection of two black areas)

3 horizontal stripe projection of the left projector,

4 horizontal stripe projection of the right projector,

5 vertical streak projection of the left projector,

6 vertical stripe projection of the right projector,

7 recognized grid and largest, common projection surface,

8th warped image of the left projector,

9 warped image of the right projector,

10 rectified image of the left projector (right projector is covered)

11 equalized image of both projectors on top of each other and

12 Comparison of luminosity.

Generation, recording and Image filtering of the stripe pattern

In order to carry out a successful geometry equalization using image warping, the underlying geometry must first be recorded. This is done by recording projected vertical and horizontal stripe patterns. These strips have a constant spacing, which can be selected depending on the complexity of the geometry to be rectified. In the exemplary embodiment described here, a strip spacing of twenty pixels was chosen because the wall had many irregularities. Other strip spacings and arrangements are possible. The projected stripes are recorded for all projectors (in our case two projectors => two vertical and two horizontal stripe images). In addition, an image is recorded in which both projectors are switched to black in order to be able to remove the partially disturbing background or low noise. This background image is subtracted from all captured images before further image filtering. The pictures taken are in the 2 to 6 to see.

Due to the use of a lens camera the recorded images show typical radial distortions, that must be removed before further processing. With distortion-free recording this step can be omitted.

The actual filtering was done with Standard image processing filter operations performed. After the image subtraction of the image to be filtered and the background image, the result image is converted into grayscale so that each of the three color channels has the same color value. This image is then converted into a black / white image depending on a threshold value, the stripes being white and the background black. To remove holes in the strips, dilation and erosion are carried out. Then the recorded lines have to be filtered so that they are only one pixel wide. This is done using a vertical or horizontal offset filter. This offset filter is carried out in two directions (left and right or top and bottom). The mean of the results of these two filtering operations provides the center of the strip. The filtered strips are now represented by their centers.

All Filter operations were implemented using hardware-supported pixel shaders.

calculation the greatest possible projection

In order to ensure congruent perception, the largest common projection area of all projectors must be determined in the original aspect ratio (4: 3 in this exemplary embodiment). To do this, a grid is first created from the filtered stripe patterns. With the help of an optimization algorithm, the largest common projection surface is "searched" within this network, which in our case consists of two sub-grids. The grating of the left and right projectors and the target projection surface are in 7 shown. The parameters for warping can then be determined from this data.

calculation the warp fields and image warping

Using the grid previously calculated the image can be broken down into warping patches. These patches exist from a Dest patch (the image area to which a part of the image was warped and an Src patch (the image area that is in the Dest patch warped).

The corner point calculation of the Dest patches is trivial. Due to the specified distance of the horizontal and vertical lines is the position of all grid points in the target known.

The Src patches are calculated linearly depending on the projection rectangle (in the photo) and the position of the Dest patches (in the photo). The calculation for the lower left corner of a Src patch is shown depending on the lower left corner of a Dest patch:
patch.x1 = (pLB.X - projectionRect.left) / (projectionRect.right - projectionRect.left);
patch.y1 = (pLB.Y - projectionRect.bottom) / (projectionRect.top - projectionRect.bottom).

The Src patch is implemented using a Warping procedure including integrated perspective correction transformed into the Dest patch.

Is the distance of the horizontal and vertical lines chosen relatively small, is also the compensation of curvatures realizable. With the line spacing of twenty pixels it was already possible curvatures balance the wall.

With a two-projector calibration, two warp arrays have to be calculated, with more than two projectors, correspondingly more. The warp arrays are always determined depending on the corresponding grid points with the limitation of the largest common projection area. The warped images calculated for a two-channel projection are in 8th and 9 to see. The result on the projection surface is in for a projector 10 and for two projectors in 11 to see. The comparison of the luminosity is in 12 shown.

To ensure effective processing, the autocalibration is carried out only once (per session). For this reason, the data calculated for the geometry equalization and multi-projector calibration are stored in warp fields, which are used, for example, by the X-Rooms ^TM system (K. Isakovic, T. Dudziak, K. Köchy. X-Rooms. A PC-based immersive visualization environment, Web 3D Conference, Tempe, Arizona, 2002) can be used. The equalization using the warp fields was again implemented with hardware support, so that equalization is possible in real time.

application areas

For There are many areas of application for multiprojector calibration. For example, it can be used to increase the luminosity or be used for stereo calibration. Furthermore, without huge Geometry and projector-related equalization can be carried out. The Autocalibration of a two-projector system using the method according to the invention takes 35 seconds including the image acquisition (approx. 20 seconds). It represents a significant increase over manual calibration represents.

Claims (1)

System for the automatic calibration of a multi-projector system with - with at least two projectors, - a digital camera and - a control unit for controlling the projectors and the camera, - an auto-calibration being carried out according to the following steps: - generation, recording and image filtering of streaks fenpattern, - Find the largest possible common projection area and - Calculate the warp fields and image warping.