Research On Low Illumination Image And Video Enhancement Algorithms Under Complex Lighting Conditions Based On Retinex Theory

Images and videos,as important carriers of information,have already spread in all aspects of human life.However,due to weather,lighting conditions,acquisition equipment and other factors,the captured images or videos are often unsatisfactory.For example,images taken in a backlit environment or surveillance videos taken at night are bright in places where there is light,and dark in places where there is no light.These unevenly illuminated images and videos,on the one hand,affect the visual experience of the human eye,on the other hand,are not conducive to subsequent applications such as computer vision.In view of the above problems,this paper analyzes the deficiency of the classic center-surround Retinex algorithm on such images based on the characteristics of low-illumination images and videos with uneven illumination,and proposes a low-illumination image enhancement algorithm with brightness perception based on Retinex theory.The algorithm is applied to video,and a flicker-free low-illuminance video enhancement algorithm framework based on Retinex theory is proposed.The specific work and innovations of this paper are as follows:(1)Aiming at the traditional Retinex enhancement with large noise and unnatural results,an improved Retinex-based image enhancement algorithm is proposed in this paper.First,the initial illumination is estimated by the method of local maximum value,and then the final illumination is estimated by the joint filtering with edge preservation to ensure that the illumination is smooth enough and the edge is clear enough,so as to obtain a reflection component with clear enough details.Because the reflection component obtained after removing all the illuminance is the reflectivity of the object,the reflectivity as a result of the enhancement is often not natural.Thus,in this paper,the adaptive weighting factor is used to control the degree of illuminance removal.The illuminance removal in the dark area is less to prevent noise amplification,and the illuminance is removed more in the bright area to obtain more details.The brightness of the reflected image obtained through the operation in the logarithmic domain is compressed in a smaller range.In this paper,Logarithmic image processing subtraction(LIPS)is used to stretch the brightness of the reflected image,in that the reflected image is similar to a dark image obtained through a thicker light absorption filter.The logarithmic image processing subtraction model is to subtract the thickness of the light absorption filter to obtain a brighter enhanced image.Experimental results show that this algorithm can better preserve the details of bright areas,and enhance the brightness of dark areas.Besides,the enhancement results are in less perceived distortion and more natural.(2)Based on the image enhancement algorithm,this paper applies it to low-illuminance videos,and proposes a flicker-free low-illuminance video enhancement algorithm based on Retinex.Since the estimated brightness of the reflected image is compressed into a smaller brightness range,the subsequent stretching of the brightness of the reflected image's LIPS parameter has a greater impact on the final enhanced brightness.In this paper,we control the value of the LIPS parameter to eliminate brightness jitter.The value is determined based on whether the scene and brightness change between two adjacent frames.The scene mutation is judged by the change of the background area of the frame.This paper uses the optical flow method to estimate the background area.In addition,the sudden change in brightness is judged by the average brightness difference between adjacent frames.If the scene and brightness of two adjacent frames are not abruptly changed,and the reflection images obtained are also similar,the LIPS parameter of the two frames must be similar to ensure that the brightness of two enhanced frames is continuous.The simple method is to use the same parameter of the previous frame to enhance the current frame.However,the accumulation of parameter often leads to the over-enhancement or under-enhancement of the frame that is far away from the first frame of the scene.Therefore,this paper adopts the method of parameter feedback adjustment.On the one hand,it avoids the problem of brightness flicker in enhanced video,on the other hand,it improves the enhancement effect of video frames.For frames with abrupt in scene or brightness,the LIPS parameter is updated through adaptive calculation.Experiments demonstrate that the algorithm in this paper can effectively improve the brightness of the video,and there is no brightness flicker in the enhanced sequence,and the overall quality of the video is improved.