You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If you extract all frames and AND their set pixels, some of noise pixels cancel themselves out. That results in a less noisy image which is easier to recognize, not harder (compared to providing a single frame, non-animated image).
To overcome this weak point, I think for each frame, we have to move the text a bit randomly. An alternative method for 'snow' would be XORing the noise pixels with the base image; i.e. if the noise pixel falls on a set pixel in the original image, unset it.
The text was updated successfully, but these errors were encountered:
it totally makes sense what you say, would you be interested in creating a patch for your suggestion? I think it would enhance the robustness against graphic-analyizing bots.
If you extract all frames and AND their set pixels, some of noise pixels cancel themselves out. That results in a less noisy image which is easier to recognize, not harder (compared to providing a single frame, non-animated image).
To overcome this weak point, I think for each frame, we have to move the text a bit randomly. An alternative method for 'snow' would be XORing the noise pixels with the base image; i.e. if the noise pixel falls on a set pixel in the original image, unset it.
The text was updated successfully, but these errors were encountered: