Research And Application Of Image Dehazing Based On Dark Primary Colors

At present,a lot of research in the field of computer vision is based on clear images collected outdoors.In actual situations,many outdoor image acquisition devices face many possibilities.For example,due to the weather,the images collected under foggy weather will seriously deteriorate The quality and definition are poor,which will bring great problems to further target recognition and detection.Therefore,in order to enhance its practicability,it is very important to realize image defogging through related technical means.For outdoor surveillance systems,real-time video defogging in foggy weather will bring better scene recognition and identify targets in the scene,so real-time video defogging has important practical significance and application.Based on the atmospheric scattering model,the defogging algorithm based on the dark primary color a priori theory analyzes the physical causes of the image degradation in foggy days,and restores the image by inverting the foggy image degradation model.The recovered images are rich in details and have good quality.Visual effect.However,the apriori theory of dark primary colors is proposed for images without sky areas,and there are still certain limitations in application.Therefore,this paper proposes improvements for the inadequacy of dark primary colors prior to defogging,mainly in the following aspects:Work: Aiming at the edge of the restored image,using the minimum image of the original image as a guide filter for the image transmittance map and the restored image;for the dark primary color theory not applicable to the sky area of the image,the Otsu algorithm and gradient feature are proposed The image is divided into two parts and combined with the median value of the neighborhood to re-estimate the transmittance of the sky area algorithm FMTO(Field median transmission optimization),and finally use the atmospheric scattering model to restore the image;for the dark channel acquisition time complexity is high It is proposed to use the block method to calculate;finally,based on the improved defogging algorithm for video defogging,a frame skipping method is proposed to speed up the video processing,and applied to the video collected by outdoor surveillance equipment to better identify the target in the video.The experimental results show that the sky segmentation scheme proposed in this paper can achieve a good defogging effect in the sky area of the restored image.The restored image has prominent edges and rich details.Its average gradient is increased by more than 12%,and the information entropy is increased by about 5%,the use of theblock method to obtain the dark channel of the image and the video defogging scheme based on the frame skip method make the original algorithm more than 3 times more efficient,and the video after defogging has a higher scene recognition.