After the correction of every image, Cosima will create (on demand) a report file, both as pure ASCII textfile and as an HTML file, which can be opened with a browser. In the report, you find at first the image errors in a small table, here is an example:
Error analysis: | ||||
Mean error of height | = | 16 | pixel | (right image upturned) |
Mean error of size | = | 0.44 | percent | (right image oversized) |
Mean error of rotation | = | 0.089 | degrees | (approx. identical rotation) |
Mean error of vergence_v | = | 0.82 | degrees | (convergent shooting axes) |
Mean error of vergence_h | = | -0.069 | degrees | (approx. equal shooting levels) |
Near point disparity | = | 8 | pixel | (near point behind stereo window) |
The last entry measures the disparity of the near point in the original image. It is negative, if the near point is placed in front of the stereo window.
With the next table, the correction values of the last iteration are given. These values normally should be very small and will converge for longer iterations against zero. If there are matching errors of corresponding image points, these values may keep large and possibly no image will be rendered. Here is an example of a successful image correction:
Image analysis after the last iteration: | ||||
Mean error of height | = | -0.13 | pixel | |
Mean error of size | = | 0.072 | percent | |
Mean error of rotation | = | 0.013 | degrees | |
Mean error of vergence_v | = | -0.0042 | degrees | |
Mean error of vergence_h | = | 0.0087 | degrees |
Because the stereo window values after the image correction are so important for the perception of a stereo image, they are summarized with a further table:
Stereo window after the correction: | ||||
Far point disparity | = | 70 | pixel | |
Near point disparity | = | 8 | pixel | |
Absolute deviation | = | 62 | pixel | |
Relative deviation | = 1/ | 28 | of the image width | |
... | = | 36 | Promille |
The absolute image deviation is the difference of the far point disparity and
the near point disparity, in this case 70 pixel - 8 pixel = 62 pixel.
To compute the relative deviation, this value is related to the image width
(in this case the image width = 1740 pixel: 62/1740 = 1/28 = 0.00356 = 36 promille).
(If the parameter ProjectionRatio specifies a fixed projection ratio,
the relative deviation is additionally given in relation to the width of the projection image.)
These stereo window values will now be tested against the following 3 tests: the deviation test, the far point test and the near point test.
If no projection ratio was specified, the deviation is based to the available image width.
However, if the parameter Projection Ratio specifies a width-to-height ratio of the projected image,
all 3 tests are performed on the base of the width of the projection image.
The 70-minute condition by Hermann Lüscher will be applied to an image,
which is observed with a presence of 0.6.
(The presence is the ratio of the projected image width and the distance to the screen,
please see
http://www.herbig-3d.de/german/minuten.htm, chapter 2.3.)
For that projection situation, deviation values larger than 1/30 of the image width
will lead to an image decomposition independent on the stereo window framing.
The deviation is definitely too large for both print and projection.If 1/20 < deviation < 1/15:
The image unsuited for publishing!
The deviation is definitely too large for projection!If 1/30 < deviation > 1/20:
The image only suited for prints!
The deviation may be critical for huge projection (depends on image content)!If 1/100 < deviation < 1/30:
The image is suited for prints!
The deviation is optimal. The image is suited for both prints and projection.If deviation < 1/100:
The deviation is too small for a typical stereo image.
The near point will be set automatically by the program or manually by the user.
Near point disparities < 0 are possible results from the automatic
only with the setting EstimateWindow = 3 and 5,
otherwise they must be produced by the user itself.
Cosima creates the following messages:
If EstimateWindow = 3, 5: Near point as designed in front of the stereo window, please check effect!
If EstimateWindow = 0, 1, 2, 4: Warning: Near point in front of the stereo window,If 0 < near point disparity < image width/50:
stereo window condition probably violated!
Near point close behind the stereo window, favourable stereo mounting.If image width/50 < near point disparity:
Warning: Near point very far behind the stereo window, unfavourable mounting!
Finally, a far point test is performed, which is extremely important, especially with near point framing.
Cosima creates the following messages :
Warning: far point disparity definitely to large, painful projection!If image width/20 < far point disparity < image width/15:
Warning: critical far point disparity, please avoid huge projection -If image width/30 < far point disparity < image width/20:
or adjust projectors to set the stereo window in front of the screen!
Warning: far point disparity may be critical (depends on image content),If far point disparity < image width/30:
please avoid huge projection -
or adjust projectors to set the stereo window in front of the screen!
Far point disparity is tolerable, the audience will be grateful.