Research And Application On Enhancement Method For Low Illumination Image

Edge detection of digital images is a fundamental problem in image processing and computer vision.It plays a key role in image segmentation,motion analysis,target recognition and tracking.Although the research related to image edge detection has a long history of development,and the existing image edge detection algorithm is also numerous in number,it can obtain sharper,more detailed saliency outlines and effectively remove noise boundaries in low illumination environments.The edge detection algorithm still has very important practical significance.In this paper,according to the characteristics of the images acquired under the low illumination environment,the existing algorithms are analyzed and enhancement method for low illumination image are improved and perfected.Then,an improved structural forest edge detector is implemented for images in low illumination environment.The main work of this paper is as follows:(1)This paper presents a low illumination image enhancement algorithm based on multi-scale Retinex theory with color restoration.After studying the traditional related low illumination image enhancement algorithm,this paper chooses the multi-scale Retinex algorithm with color recovery to estimate the illumination component and reflection component of the image.Then the gamma transform and the linear stretching process are performed on the illumination components,and the reflection components are filtered by trilateral filter.Finally,automatic white balance processing is performed on the combined image of the illumination and reflection component.Through the analysis of the experimental data,compared with the traditional algorithm,the proposed algorithm is more subjective,the details are more prominent,and the overall brightness is improved significantly.In the objective aspect,the experimental data show the superiority of the algorithm.(2)An improved edge detector based on structured forest.After studying the traditional related edge detection algorithms,this paper uses the structured random forests as a classifier and uses the Berkeley segmentation data set(BSDS500)to train the structured random forest algorithm.The core is to structure the given nodes.Tags are mapped to a set of discrete tags.Then using the trained random forest algorithm to classify enhancement method for low illumination image processed image information to obtain the image outline.Finally,Otsu is used to perform automatic thresholding on the contour image to obtain the prominent contour of the image.Experimental results show that the proposed algorithm can obtain sharper and more detailed salient contours and effectively remove noise boundaries,effectively improving the detection efficiency.(3)Realized the system application of enhancement method for low illumination image and edge detection algorithm proposed in this paper.This article will deal with low-light image enhancement algorithms for real-time streaming video frames acquired from IP network cameras,combining structural forest edge detectors with directed watershed computation and UCM algorithms for image segmentation.Experimental results show that the low-light image edge detection algorithm based on structured forest proposed in this paper has a certain value in practical application.