Research On Nighttime Video Image Enhancement Based On Retinex Theory

In real life,due to the influence of illumination conditions,the boundary and the details of some images are blurred.And the recognition rate is low.Thus,in the condition of keeping video images in lower brightness,how to improve recognition of video images has also become a new direction of image processing and image enhancement.In the research of this paper,it talks about the image enhancement in low luminance condition which based on the Retinex theory.When analyse the SSR and MSR Retinex image enhancement algorithm,there are some problems,including halo in the Retinex image enhancement algorithm,noise is magnified in the original images and lower Retinex iterative efficiency.To solving the problems in Retinex image enhancement algorithm,this paper come up with some ideas,like the variable frame Retinex algorithm and reducing the smoothness of the image of the incident light.Thus,better image edge information and detail information are retained.And the halo of Retinex algorithm can be eliminated effectively.But when retaining the image edge information and detail information with modified Retinex algorithm,the noise could also be enlarged.For this reason,this paper talks about how to use a bilateral filter to eliminate the enlarged noise information of the improved Retinex image enhancement algorithm.when iterating to obtain the approximate optimal value of the solution in the Retinex image enhancement technology,a large number of iterations inevitably led to the image enhancement algorithm efficiency decreased.Therefore,in this research,we can improve the efficiency of the algorithm by combining the convolution and the multi scale process.By the improved algorithm,the contrast experiment of SSR and MSR algorithm,the improved algorithm on the three components of RGB image processing won better treatment effect.After using the improved algorithm,higher mean image and standard deviation appeared.It shows that the effect of improved algorithm of image enhancement is better if there is no increase of brightness of the image.While improved algorithm of information entropy is minimal.It indicates that the improved algorithm has the least impact on noise.So image information recognition efficiency is higher in it.