Abstract: For the salt-pepper noise which deduced in the gaining, processing, transmission of the visual text images, a new algorithm for removal of salt-pepper noises is proposed, which based on the characters stroke characteristic. This algorithm could estimate whether a pixel and it’s neighboring pixels are salt-pepper noises or not, and differentiate effectively characters from salt-pepper noise. The experimental results show clearly that this algorithm could restrain salt-pepper noises well and truly, at the same time, comparing with those traditional denoise algorithms, it avoid the influence for characters stroke, and decrease the disturbance for subsequent recognition.