[esimov]

This PR is a major addition to the library and its main purpose is to address a feature request opened a while back mentioned in this issue #81, to facilitate text recognition inside an image in order to avoid the seam detector to remove undesired areas. The opted approach is not the traditional OCR text detection, but a much simpler version based on this article https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/201020CVPR20TextDetection.pdf.

We are not interested to correctly recognize the letters from an image, but more likely just to detect letters bounded to the same text group. By correctly detecting letter components we can "trick" the seam (Sobel) detector to consider them as equally important image areas as the ones detected by the Sobel operator.

Other than the basic implementation of SWT, it contains a lot of other improvements, like:

• fixed image operation regression introduced with the latest commits
• simplified and more accurate seam GUI display (was detectable by displaced seams, happened mostly on vertical image resize on certain image aspect ratio)
• fixed GIO issue on centering image window to the center of the screen (resolved with version upgrade)
• other minor adjustments and code refactoring