Hardware for image alignment by one shear transformation
Wan-Soo Choi, Yun-Koo Chung · 2005
This paper presents a hardware for image alignment in an OCR system. This hardware uses a new image-aligning algorithm which performs one shear transformation with no losing pixel information when compared with traditional aligning algorithms. One shear transformation is performed by simply shifting pixels into the target area along a base line without computing new positions of all pixels for aligning. The effectiveness of this algorithm is demonstrated by aligning algorithms. In the hardware, a misaligned image made up of 480 by 380 pixels is aligned within 65 ms.