手写体数字识别中一种新的倾斜校正的方法

A New Method for Slant Correction in Recognition of Handwritten Numerals

介绍了在手写体数字识别预处理过程中一种新的倾斜校正的方法,该方法利用手写数字的图像的高与宽的比值是否最大来确定倾斜校正是否完毕。在校正过程中,手写体数字的图像高度逐渐变高,宽度逐渐变窄,所以当图像的高与宽的比值最大的时候倾斜角度最小。

A new method for correcting askew handwritten numerals in the preprocess of recognition is presented in the paper. In order to know that the correcting process has ended, it should make sure that the ratio between the height and the width of the bitmap of the handwritten numerals is the greatest. In the process of slant correction, the height of the bitmap is increasing, while the width of the bitmap is decreasing, so the bitmap is least slanted when the ratio between the height and the width becomes the greatest.