A Low-Illumination Color Video Enhancement Method Using Spatio-Temporal Illuminance Map

Abstract

The invention discloses a low-illuminance color video enhancement method using a spatio-temporal illuminance map. Firstly, the initial spatial illuminance map of the video frame is calculated, and the temporal similarity of the current video frame is calculated by using the initial spatial illuminance map of the current video frame and its previous k video frames. Illumination map, using the initial spatial illuminance map and temporal similarity illuminance map to construct the structural weight matrix of the current video frame, using the initial spatial illuminance map, temporal similarity illuminance map and structural weight matrix to construct a cost minimization model for optimal illuminance map estimation and Solve the solution, the obtained solution is the optimal illuminance map of the current video frame, and finally perform gamma correction on the optimal illuminance map, and use the corrected illuminance map to obtain the enhanced current video frame; when all video frames are processed, the All enhanced video frames are merged into an output color video. The invention can effectively enhance the illuminance of low-illuminance color video while maintaining the main structure in the video frame, and makes the enhanced video have coherent illuminance.

Description

technical field [0001] The invention relates to a low-illuminance color video enhancement method using a spatio-temporal illuminance map, which belongs to the image enhancement technology. Background technique [0002] In the case of insufficient natural light, the videos captured by people often have low illumination and poor visual effects, which cannot truly reflect the details of objects in the target scene. For some vision-based applications, such as target detection, target Tracking, etc., have caused great inconvenience. [0003] Methods for enhancing low-light images or videos can be roughly divided into two categories, one is based on the histogram, and the other is based on the Retinex theory. The histogram-based method improves the resulting image or video frame by adjusting the histogram distribution of the image or video frame. However, it does not solve the problem of illumination enhancement from the essence of image illumination, but only improves the image ...

AI Technical Summary

Problems solved by technology

A fusion-based enhancing method for weakly illuminated images, vol.129, no. 2016, pp.82-96, 2016), three different illumination maps are generated in three ways, and then weighted and fused into an enhanced illumination map. However, this method does not consider the structural information in the image, which will cause some textured areas lighting distortion

The Chinese patent "A Low Illumination Video Image Enhancement Method" [201310639998.5] performs intra-frame enhancement on the illumination of each frame of image, and performs inter-frame enhancement on the previous frame enhancement result as the standard frame of the next frame image. However, this method also does not take into account the structural information in the video frame

Chinese patent "A Real-time Enhancement Method for Low-Illumination Video Based on Bright Channel" [201510181696.7] obtains an illumination factor by calculating the mean value of low-illuminance video frames and the value of the bright channel, and the simple relationship between the illumination factor and the pixel intensity of the video frame The algebraic operation of the method is used to obtain the enhanced video frame. This method is fast and can meet the requirements of real-time processing, but it does not use the inter-frame information of the video.

Xiaojie Guo et al proposed a low-light image enhancement method (Xiaojie Guo, Yu Li, Haibin Ling, LIME: Low-light image enhancement via illumination map estimation, IEEE Transactions on Image Processing, vol.26, no.2, pp.982- 993, 2017), by estimating the illuminance map to achieve the purpose of enhancing the image illuminance, this method takes into account the spatial structure information inside the image, and can estimate a spatially segmented and smooth illuminance map, but it is not suitable for processing low-level images that contain temporal information. Illumination video

Images