BCSR: toward arbitrarily oriented text image super-resolution via adaptive Bezier curve network

Abstract Although existing super-resolution networks based on deep learning have obtained good results, it is still challenging to achieve an ideal visual effect for irregular texts, especially spatially deformed ones. In this paper, we propose a robust Bezier Curve-based image network (BCSR), which can efficiently handle the degradation caused by deformations. Firstly, arbitrarily shaped text adaptively fitted parameterized curve, aiming convert curved box into annotated box. Then, design BezierAlign layer calibrate between extracted features and input image. By importing prior information, accuracy of be significantly improved. It worth highlighting that kind loss enables cooperation enhancement. Extensive experiments several standard scene datasets demonstrate our proposed model achieves desirable objective evaluation results further immensely helps downstream tasks related recognition, in instances with multi-orientation shapes.