Research On Automobile Anti-halo Method Based On Infrared And Visible Image Fusion

As a commonly used means of transportation,with the continuous improvement of people’s living standards,the global car ownership has also increased year by year,causing more and more traffic accidents.One of the reasons is the improper use of the high beams of the driving at night,especially when meeting cars at night,the unreasonable use of the high beams of the oncoming vehicle will cause the driver to produce instant visual blind spots,which will cause great Hidden dangers of driving safety.On the other hand,due to the impact of the halo of vehicles driving at night,once a traffic accident escapes,a single imaging camera cannot capture the effective information of the vehicle,resulting in insufficient evidence of the traffic accident.Aiming at the above problems,this thesis mainly conducts anti-blooming research from two aspects: the denoising method of the halo image and the image fusion method.In terms of image denoising,the traditional median filter cannot accurately distinguish the noise points in the halo area wen removing the halo image noise,resulting in image distortion after denoising and reduced contrast.This thesis proposes an improved image denoising method based on adaptive fuzzy median filtering.The entire filtering process is divided into three parts: coarse noise detection,fine noise detection,and noise filtering.And this thesis mainly introduces the average gradient difference average value in the noise fine detection,and constructs the fuzzy variable by using the absolute average gradient difference average value of the suspected noise points obtained from the rough detection and the signal points in the window(including the filtered noise points).Variables perform secondary detection on suspected noise points to determine whether they are noise and perform fuzzy classification processing;and then perform different filtering processing on the results of noise fuzzy classification.Experimental simulation results show that the method proposed in this thesis has a significant improvement in PSNR(peak signal to noise ratio),MSE(mean square error),and SSIM(structure similarity)compared with the traditional median filtering algorithm.In terms of image fusion,the current traditional infrared and visible light image fusion methods can eliminate car halo while causing the loss of image texture details and the blurred information of vehicle characters.This thesis proposes an improved non-subsampled contourlet transform(Non-subsampled Contourlet Transform(NSCT)adaptive image fusion method.First,the registered image is multi-scaled and decomposed into high and low frequency components by NSCT transform.Aiming at the pixel characteristics of low-frequency components,this thesis proposes a fusion method based on regional variance adaptive weighting.For high-frequency components,this thesis proposes a fusion strategy that combines the large average gradient and the large absolute value of the pixel.The experimental results show that the fusion method proposed in this thesis can effectively eliminate car halo and keep the image texture details unaffected,highlighting the target information.