There is a video stream . That is, infinite .

In the video there may be places that need to be blurred. Well, for example, the inscription "go all on ..." on the fence.

A handwritten program should analyze the video flow that it receives at the input, and if there is a fence and such an inscription, blur the inscription. And issue the same video with a blurred caption.

Or just a video without changes, if there is not this.

That is, as a filter.

How about this can be done? where to dig?

  • In such a formulation - no way. If the video is pre-marked and in parallel with the stream, there is some kind of meta-information about what coordinates you need to do the blurring, then something else can be invented. - Xander
  • @ Alexander, well, why is it possible to recognize? to train. - Jodimoro
  • take a neural network for text recognition, sranivayte text with a template, and if converge blurs. Problems? - Andrio Skur
  • one
    @Alexander Real Time Text-Recognition. (OpenCV) - jfs

0