Stroke-based Character Reconstruction
This addresses a challenging issue in character recognition for applications like document analysis or scene text processing, but it appears incremental as it builds on existing stroke-based approaches.
The paper tackles the problem of background elimination in noisy or real-scene character images by proposing a stroke-based character reconstruction method using weighted quadratic Bezier curves, achieving excellent reconstruction in real scenes and strong defense against adversarial attacks on character recognizers.
Background elimination for noisy character images or character images from real scene is still a challenging problem, due to the bewildering backgrounds, uneven illumination, low resolution and different distortions. We propose a stroke-based character reconstruction(SCR) method that use a weighted quadratic Bezier curve(WQBC) to represent strokes of a character. Only training on our synthetic data, our stroke extractor can achieve excellent reconstruction effect in real scenes. Meanwhile. It can also help achieve great ability in defending adversarial attacks of character recognizers.