[Blog Post] MSI at last!

[oshfdk]

I find it a bit hard on the eyes to look at the processed images from MSI, so I tried to use a support vector classifier (SVC) trained on the pixel values of the visible text on the margin of 1r (from MSI TIFFs) to create a BW image of the whole page. There is really nothing here that cannot be seen in the original TIFF images, but it is much easier to view. Since it was trained on the margin, the model couldn't separate the main text on the page from the drawing on 1v.

One curious thing (I don't remember if this was mentioned before): there is something that looks like a Latin 'a' at the tip of EVA 'ees'.

   

   

Here's the full resolution image of the whole page (5MB PNG): 
You are not allowed to view links. Register or Login to view.