A new method for converting trimmed NURBS surface to T-splinesurface for isogeometric analysis AIP Conference Proceedings 1834, 040029 (2017) Text images captured by camera may be tilted and distorted, which is unfavorable for document character recognition.Therefore,a method of text image tilt correction based on wavelet pyramid is proposed in this paper. The first step is to convert the text image captured by cameras to binary images.After binarization, the images are layered by wavelet transform to achieve noise reduction, enhancement and compression of image.Afterwards,the image would bedetected for edge by Canny operator, and extracted for straight lines by Radon transform. In the final step, this method calculates the intersection of straight lines and gets the corrected text images according to the intersection points and perspective transformation. The experimental result shows this method can correct text images accurately.