书籍扫描图像畸变模型的可视化描述及应用

利用设备无关显示技术在计算机屏幕上显示一条仿真曲线、一个矩形和一个箭头,可视化地描述在扫描时书籍页边的变形情况、页面上的均匀条带及成像系统对称中心的位置。通过调整仿真曲线,使之较好地反映页边实际变形,并使矩形的大小和位置及箭头的位置合适,可对存在严重畸变的书籍扫描图像进行较好的校正,而无需用户精确地测量这些随具体情况而变化的参数。校正后,用OCR(Optical Character Recognition)软件进行文字识别时,总的平均纠错率为94.5%,页平均纠错率最高达99.6%。用TWAIN(Technology Without An Interesting Name)标准直接从扫描仪读取图像数据,并用椭圆拟合方法方便地测量扫描仪的成像畸变系数。

Visualization of the distortion model for scanned book image and its application

Using device-independent displaying technology, a curve was displayed on computer screen to describe the twisted page edge of book that was played on scanner, a rectangle was displayed to select a band within witch there was nothing will cause the image gray scale change acutely, and an arrow to indicate the scanner's optical system center. The curve was adjusted by operators to make it look like the twisted page edge exactly from shape to size. The rectangle's position and size was made suitable, also was the arrow's position. Then the parameters of the curve, the rectangle and the arrow were used to correct the image distortion of that page, and obtained very satisfying effect without measuring accurately these parameters that will change frequently under different conditions. Before correction, many characters could not be recognized by OCR programs. After correction, 94.5 total average percent of the can-not- recognized characters was recognized, for single page the best average percent was 99.6. TWAIN standard was used to read image data from scanner directly, and ellipse curve fitting was used to determine the distortion parameter of the scanner's optical system.