0

I have an image. I resized it i obtain another with a size greater than the first .

    resize(roi,zoom,Size(2274,70),(0.0),(0.0),3);

i also test all the method of interpplation but it don't give a good result.

     CV_INTER_NN (default)
     CV_INTER_LINEAR
     CV_INTER_CUBIC
     CV_INTER_AREA

The image contain text when i zoomed it it becomes so fuzzy and i can not recognize the text.

I ask for an algorithm or method to make the image more clear

Thanks for the help

user3770614
  • 25
  • 1
  • 9
  • You could provide links to the original *and* the corresponding result images in order to make sure we understand your situation correctly. – Reunanen Jun 27 '14 at 15:25
  • If your original image is just slightly smaller than the result that you linked, this is an expected result and it's very hard to write an algorithm that would work well in such a situation. Essentially, you should not scale an image but rather render a new one with a larger font and so on. – Reunanen Jun 27 '14 at 15:28
  • 1
    I cannot recognize the text in the original image either, at least not any better than in the zoomed version. I don't think any zooming algorithm can find detail that does not exist in the original image. I am afraid that you need to carefully re-think what it is that you want to accomplish. – Reunanen Jun 27 '14 at 15:31
  • @Pukku do you know what super resolution do in opencv ? – user3770614 Jun 27 '14 at 15:40
  • No, not really. But given e.g. this paper [http://www3.nd.edu/~jsimpkin/Citations/simpkinsSRCh2012.pdf], I believe you are supposed to have *a set* of input images. – Reunanen Jun 27 '14 at 15:43
  • Also if you have a suitable prior for a specific purpose, you may be able to do magic. But in general from a single image you can't recover information that simply is not there. – Reunanen Jun 27 '14 at 15:45
  • like pukku said, you can't recover data that was never there. maybe you could try to sharpen the image? try this:http://stackoverflow.com/questions/4993082/how-to-sharpen-an-image-in-opencv – iedoc Jun 27 '14 at 15:47

0 Answers0