Diffevent: Event Residual Diffusion for Image Deblurring

Traditional frame-based cameras inevitably suffer from non-uniform blur in real-world scenarios. Event cameras that record the intensity changes with high temporal resolution provide an effective solution for image deblurring. In this paper, we formulate the event-based image deblurring as an image generation problem by designing diffusion priors for the image and residual. Specifically, we propose an alternative diffusion sampling framework to jointly estimate clear and residual images to ensure the quality of the final result. In addition, to further enhance the subtle details, a pseudoinverse guidance module is leveraged to guide the prediction closer to the input with event data. Note that the proposed method can effectively handle the real unknown degradation without kernel estimation. The experiments on the benchmark event datasets demonstrate the effectiveness of our method.

1. INTRODUCTION

Traditional cameras integrate scene information over the exposure time. Blurring frequently occurs in captured pictures due to motion between scenes and the handheld camera. This blurring causes finer details in the scene to be lost, especially in cases of non-uniform blur, making it challenging for most existing deblurring methods [1], [2], [3], [4] that rely on the blurred image to achieve practical and ideal results in real-world scenarios. Event cameras, instead of solely providing visible images like traditional cameras, detect changes in scene radiance intensity. They transmit a sequence of asynchronous event streams with low latency and high temporal resolution, providing available motion information during exposure. Therefore, the primary focus of our work is to enhance deblurring performance within the framework of event cameras.