Merging two Different Exposed Images

The composition of the resulting image from an under- and overexposed image consists of three types of pixels. Pixels around the welding process are supplied directly from the unde-rexposed image. In figure 4.26 these pixels are marked with a red colour. The background pixels are taken directly from the overexposed image. They are marked in figure 4.26 with a blue colour. These pixels do not contain any information in the underexposed image as these pixels are clipped to black. Some regions are available in both images. Following the merging algorithm of section 3.1.4 they are merged by using the mean value from both images and are marked with a green colour.

4.1. Pulsed MAG process

Figure 4.26.: MAG process at 85 A: Merge composition of the resulting image from the underexposed (red), both (green) and overexposed (blue) image

Merging the images results in data with exceeding the value range of a standard grey scale image of a computer. More than 256 different grey scales values are not available on the screen and nor the human eye can perceive all. One way to make it visible is to use false colours which map a high amount of different grey scale values to a colour map. The colour map is deviated from the hue-saturation-value colour model, where the hue component is varied (see [74]). The colour map begins with red, passes through yellow, green, cyan, blue, magenta and returns to red.

MAG, 85 A In figure 4.27 such a mapping is made to show the enhanced information contained in the merging result. The magnified area of the welding process shows that the image information contains a higher number of different grey values. If only the data taken from the underexposed image is taken into account (marked with red in figure 4.26) then 98 different grey values form the information of the welding process. By contrast the amount of an adequate area from a direct single shot as presented in section 4.1.1, covers 62 grey values. This measurement can only be used qualitatively as the region borders of the welding process are not clearly defined in the single shot image.

The resulting grey scale image in figure 4.28 is the linear transformation of the false colour image (figure 4.27) with 389 grey values to a 256 greyscale image. The transformation is equal to a histogram stretching where it is used for image improvement with the only difference that it is now done from a large data domain to a smaller one.

By applying the VBSAHEalgorithm the source image needs to be segmented before enhan-cing. The result of the detailed segmentation of figure 4.28 is shown in figure 4.29. It gives an impression about the amount and individual shape of the regions. After enhancing, the merged regions form a new image enriched in details as given in figure 4.30.

Figure 4.27.: MAG, 85 A: False Colour of the merged images with 389 greyscale values

Figure 4.28.: MAG, 85 A: Result of the merging process

4.1. Pulsed MAG process

Figure 4.29.: Segmentation of the merged image result

Figure 4.30.: MAG, 85 A: Result of the VBSAHE processing

MAG, 120 A For the process with 120 A welding current, the image composition from the underexposed image does use more regions. Not only the welding process supplies data in the underexposed image. Some bright direct lighted spots as seen in figure 4.31 are used as well as the welding arc region.

Figure 4.31.: MAG process at 120 A: Merge composition of the resulting image from the underexposed (red), both (green) and overexposed (blue) image

The composition results in an image using 337 different greyscale values and so giving more information at the welding spot, than single shot images (see figure 4.32).

By linear tone mapping the image in figure 4.33 to 255 greyscale values the result appears rather dark. It can be made re-visible by theVBSAHEalgorithm (see figure 4.34).

Figure 4.32.: MAG, 120 A: False Colour of the merged image with 337 greyscale values

Figure 4.33.:MAG, 120 A: Result of the merging process

Figure 4.34.:MAG, 120 A: Result of the VBSAHE algorithm

4.1. Pulsed MAG process

MAG, 220 A Similar to the example corresponding to 85 A welding current is the result for the brighter welding current corresponding to 220 A. The composition in figure 4.35 is as well done by a small spot of the welding arc from the underexposed image (red), an huge area where the mean value of the both input images is taken (green) and parts of the dark regions taken from the overexposed image (blue).

Figure 4.35.: MAG process at 220 A: Merge composition of the resulting image from the underexposed (red), both (green) and overexposed (blue) image

The process with its welding arc and melting pool covers at 220 A an amount of 37 grey values. This lower amount of different grey values fits to the technical specifications of the used C-MOS camera sensor. As described in section 2.2, its logarithmic response does map the same absolute difference at low lighting to a higher amount of grey value range than for the same absolute difference at bright lighting. This means that the process with its clearly bordered region is mapped to a smaller range of grey values which can be seen in the magnified region of figure 4.36.

The enhancement with the VBSAHEalgorithm of the merged and tone mapped image (see figure 4.37) supplies again a better view onto the surrounding (see figure 4.38). In the process area no specific details are visible, as they do not exist in the source image. An interesting improvement occurs in the area of the smoke on the left side of the welding torch. By that the background in that area becomes more visible. The huge global brightness differences related to the white stripe area are lowered by the VBSAHEalgorithm.

Figure 4.36.: MAG, 220 A: False Colour of the merged image with 321 greyscale values

Figure 4.37.:MAG, 220 A: Result of the merging process

Figure 4.38.:MAG, 220 A: Result of ap-plying the VBSAHE algo-rithm