Novel rain removal method and system for dynamic vision sensor event streams

Abstract

The invention relates to a novel rain removal method for dynamic vision sensor event streams. Based on the sparsity of rainwater traces in the time direction and the horizontal width direction, a traditional HW visual angle is converted into a WT visual angle, rainwater is extracted from an intricate background environment through the conversion and is changed into noise points which are approximately and uniformly distributed on a WT plane, and the rainwater removal difficulty is greatly reduced; rainwater traces are removed by using a median filtering algorithm, redundant information in theimage is fully utilized, and detail features of the image can be kept to the greatest extent while denoising is performed. The essence of median filtering is that the pixel value of a certain point isreplaced by the median of each point value in a neighborhood of the point, so that the surrounding pixel values are close to the true value, and isolated noise points are eliminated. Rainwater in theWT space is represented as salt and pepper noise, and the median filtering can simply and efficiently filter the noise and protect the sharp edge of the image.

Description

technical field [0001] The present invention relates to the technical fields of image restoration and image denoising, and more specifically, relates to a novel rain removal method and system for dynamic visual sensor event streams. Background technique [0002] Dynamic vision sensor, namely Dynamic Vision Sensor, is a special sensor based on event and applying the concept of bionics. Because of the clear layering of retinal cells and different visual channels, human beings have ultra-high resolution, extremely fast response speed, strong parallel processing capabilities, and very low power consumption. They can extract feature information in the scene instead of Simply perform absolute numerical presentation of light intensity just like ordinary traditional sensors. [0003] The demand for rain removal technology comes from the problems encountered by outdoor visual detection systems. At present, outdoor visual detection systems are widely used in object detection, data co...

AI Technical Summary

Problems solved by technology

But the present deraining technology overwhelming majority is designed for the photo or the video that traditional camera is taken, is not suitable for the video or the picture that DVS is taken; blank state, and the application of dynamic visual sensors in the field of visual detection is becoming more and more extensive. Therefore, it is of great significance and use value to study the rain removal algorithm for dynamic visual sensors

[0004] patent CN106056544A, the publication date is 2016.10.26; it discloses a video image raindrop removal method and system, which can effectively establish a video rain streak noise model and improve However, the multi-scale hybrid model of the invention is based on the scale of traditional RGB video images for processing, and is not suitable for event stream video images generated by DVS dynamic visual sensors

Most of the existing deraining algorithms are based on RGB images or videos generated by traditional cameras, and the images are processed in two dimensions of height-width (H-W); while the DVS dynamic vision sensor monitors each pixel When the light intensity changes, the pulse signal (event) will be output when it exceeds the set threshold, so the output of DVS is a binary image. The traditional deraining algorithm cannot process the event stream video generated by DVS in the height-width (H-W) Two dimensions are processed

Images