STRIVE: Scene Text Replacement In Videos

TL;DR

This paper introduces a novel method for replacing scene text in videos by combining deep style transfer, learned photometric transformations, and a three-step process to ensure realistic, temporally consistent results across challenging conditions.

Contribution

It extends still image text replacement techniques to videos, addressing challenges like lighting, motion blur, and temporal consistency with a new three-step approach and datasets.

Findings

Realistic text transfer in synthetic and real videos

Competitive quantitative and qualitative performance

Faster inference speed than existing methods

Abstract

We propose replacing scene text in videos using deep style transfer and learned photometric transformations.Building on recent progress on still image text replacement,we present extensions that alter text while preserving the appearance and motion characteristics of the original video.Compared to the problem of still image text replacement,our method addresses additional challenges introduced by video, namely effects induced by changing lighting, motion blur, diverse variations in camera-object pose over time,and preservation of temporal consistency. We parse the problem into three steps. First, the text in all frames is normalized to a frontal pose using a spatio-temporal trans-former network. Second, the text is replaced in a single reference frame using a state-of-art still-image text replacement method. Finally, the new text is transferred from the reference to remaining frames using a novel learned image transformation network that captures lighting and blur effects in a temporally consistent manner. Results on synthetic and challenging real videos show realistic text trans-fer, competitive quantitative and qualitative performance,and superior inference speed relative to alternatives. We introduce new synthetic and real-world datasets with paired text objects. To the best of our knowledge this is the first attempt at deep video text replacement.