renweizhukov / ocr

Use Tesseract OCR to recognize English and Simplified-Chinese characters.

ocr

Use Tesseract OCR or other computer vision methods to recognize English and Simplified-Chinese characters.

1. ocr-preprocessing

This exectuable preprocesses a colored image for the later OCR. After it loads the colored image, it

• sharpens the image using Unsharp Masking with a Gaussian blurred version of the image;
• converts the image into a grayscale one;
• converts the grayscale image into a binary one by thresholding.

At the end, it will write the binary thresholded image into a file.

Below is a sample usage of the exectuable.

$ ./ocr-preprocessing ./color-image.png ./binary-thresholded-image.png

2. extract-booktitle-batch

This executable extracts the title part from a set of book cover images based on a given title template image. After preprocessing the images, it will use either SURF+homography+perspectiveTransform or template matching to find the rectangle region inside the book cover images which contains the title and crop the title from the book cover images. At the end, it will write the cropped images into files.

Below is two sample usages of the executable.

$ ./extract-booktitle-batch -i title-template.png -d ./book-cover-imgs/ -o ./cropped-imgs/ -m homo
$ ./extract-booktitle-batch -i title-template.png -d ./book-cover-imgs/ -o ./cropped-imgs/ -m templ

3. ocr-circled-digits-batch

This executable recognizes the circled digits in the cover images from a series of books. After sharpening the images using Unsharp Masking with a Gaussian blurred version of the images, it will use one of the following three methods

• SURF+homography+perspectiveTransform
• Template matching
• Hough circle transform

to find the rectangle region which contains the circled digits, crop the region from the book cover images, and convert the cropped image into a black-white one via thresholding. Then it will use template matching to recognize the circled digits.

Below is three sample usages of the executable.

$ ./ocr-circled-digits-batch -i series-title.png -d ./book-cover-imgs/ -t ./digit-template-imgs/ -o ./output/homo
$ ./ocr-circled-digits-batch -i series-title.png -d ./book-cover-imgs/ -t ./digit-template-imgs/ -o ./output/templ -m templ
$ ./ocr-circled-digits-batch -i series-title.png -d ./book-cover-imgs/ -t ./digit-template-imgs/ -o ./output/hough -m hough