Settings

There are several settings that have a high impact on the results of the library.

Given this master image:

...and this test image:

When the comparison takes place, the first output is a diff. This can be done with the full colour image or a greyscale image, and is determined by calling one of the following:

    vz::ImgCmp image_compare;
//  image_compare.set_flag(vz::ImgCmp::Flags::kDiffColour   );
    image_compare.set_flag(vz::ImgCmp::Flags::kDiffGreyscale);

The colour differences look like this:

...while the equivalent greyscale differences look like this:

These images are available via vz::ImgCmp::differences.

The differences are then converted to binary masks. There are several techniques and settings involved in the process, the key ones being:

• thesholds
• dilation and erosion
• minimum area

This results in black-and-white single channel images like this:

The single-channel threshold mask is then used to obtain the contours of all the differences. A contour is a vector of 2-D points (X and Y coordinate pairs) which indentify the perimeter of each difference.

Unlike everything else so far in this process, this is not a "visual" thing, but if needed all the contours are stored together in a vector called vz::ImgCmp::differences_contours.

The last step is to draw these contours on an image. This is done in vz::ImgCmp::annotate(). There are several options that control how the contours are displayed.

Key settings are:

• vz::ImgCmp::kAnnotateOverGrey
• vz::ImgCmp::kAnnotateOverColour
• vz::ImgCmp::kAnnotateAddRedBorder
• vz::ImgCmp::kDrawRectangle
• vz::ImgCmp::kDrawContour