AnimatedCaptcha

AnimatedCaptcha copied to clipboard

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.

Hi Majid,

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.

Michael