Real-time text detection method based on differentiable binarization

Abstract

The invention discloses a real-time text detection method based on differentiable binarization. An image is segmented to obtain a probability graph of a text region, differentiable binarization is performed on the probability graph to obtain a binary image, and a connected region is searched on the binary image to obtain a bounding box of the text region. According to the method, the probability graph and the binary graph are supervised, the binarization process is brought into training, and the detection effect is improved. Compared with an existing text detection method, the method has the advantages that excellent effects are achieved in the aspects of accuracy, operation efficiency and universality, and the method has very high practical application value.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and more specifically relates to a real-time text detection method based on differentiable binarization. Background technique [0002] Reading text in scene images has become an active research area in recent years due to a wide range of practical applications in image / video understanding, visual search, autonomous driving, and blind assistance. [0003] As a key component of scene text reading, scene text detection, which aims to locate the bounding box or region of each text instance, remains a challenging task because scene text usually has various scales and shapes, including horizontal, multi- Orient and bend text. Segmentation-based scene text detection methods have attracted a lot of attention recently because it can describe texts of various shapes benefiting from its prediction results at the pixel level. However, most segmentation-based methods require complex post-processing ...

AI Technical Summary

Problems solved by technology

However, the standard binarization function is not differentiable, so it cann...

Images