Calling swtdetect (which uses ccv_swt_detect_words to detect words) on a scanned image from a book, and postprocessing result to red box each "word" detected, with default params I get this result:
Input is a 70 MB TIFF image (@ 600 x 600 dpi, 6992 x 4920 pixels).
As you can see it almost perfect, but it has some true negatives and some false positives.
Since my goal is to detect page external "margins"(top, right, bottom, left areas with no text), the result is spoiled by the two bottom false positives.
Is there any documentation about ccv_swt_default_params fields, to avoid these false postives (and possibly decrease true negatives, too)?
Calling swtdetect (which uses
ccv_swt_detect_words
to detect words) on a scanned image from a book, and postprocessing result to red box each "word" detected, with default params I get this result:Input is a 70 MB TIFF image (@ 600 x 600 dpi, 6992 x 4920 pixels).
As you can see it almost perfect, but it has some true negatives and some false positives.
Since my goal is to detect page external "margins"(top, right, bottom, left areas with no text), the result is spoiled by the two bottom false positives.
Is there any documentation about
ccv_swt_default_params
fields, to avoid these false postives (and possibly decrease true negatives, too)?