Different from the earlier user assistance that requires users to clearly input various complex information such as shadow contour, at present, some user assistance methods only need to input a small amount of information to obtain good results. User assistance shadow removal can be completed under two conditions: automatic image shadow detection and input of some image information with user assistance. Deep learning models make the original feature engineering no longer necessary, but these algorithms are often based on a large amount of data and need a long time for model training. With the development of deep learning algorithms in recent years, deep learning models have also been applied to shadow detection. At the same time, for the models that distinguish each pixel separately, most of the works used CRF, MRF, and other similar models to maintain regional consistency. Machine Learning MethodĬurrently, a lot of works combined the direct distinction of original features with machine learning algorithms to train classifiers to distinguish shadow regions under the premise of data-driven. Some color-related features were mentioned in to distinguish shadow regions. proposed the characteristics of brightness and change rate.
proposed a single feature threshold method for directly distinguishing shadow regions. emphasized the local features of the image and determined the shadow boundary quickly by comparing the features of adjacent regions. proposed a variety of features that change and remain unchanged under the shadow in the gray image, including the features of distinguishing dark objects from shadow regions. Shadow FeatureĪt present, shadow detection based on image features is widely studied, mainly to extract some changing and invariant features in shadow. At the same time, there are many ways to combine these directions. Related WorksĪt present, the commonly used shadow detection and removal methods are divided into three directions: the method based on a physical model, the method based on basic image features, and the method based on machine learning. The experiments and experimental results’ analysis are described in Section 4, and finally, this paper is briefly summarized in Section 5. In Section 3, we present the processes of shadow detection and removal. Section 2 introduces the related works of our proposed approach. The remainder of this paper is organized as follows. Other features in the shadow region can be kept unchanged to a great extent, and only the shadow is restored. (2) In shadow removal, the shadow region is restored according to the region matching. (1) In shadow detection, based on the mutual restriction between image region features and matching region, this method can detect directly without training. The contributions of this paper are as follows. This method can accurately detect the shadow region and minimize the impact of shadow removal on other features in the shadow region. Inspired by the relevant articles on shadow detection and removal through region matching and other methods, this paper proposes a shadow detection and removal method based on interarea material matching without training. The difficulty of shadow removal is to ensure that only the shadow is removed without changing other image features in the shadow region. The difficulty of shadow detection lies in the complexity of background and light in the image, as well as the interference of dark objects in the image. Therefore, the detection and removal of shadow have been widely concerned.
IntroductionĪlthough the shadow in the image is a kind of image information, it often interferes with object recognition, image segmentation, and other image processing work.
The experiments on the benchmark dataset demonstrate that the proposed approach achieves a promising performance, and its improvement is more than 6% in comparison with several advanced shadow detection methods.
In shadow removal, the proposed method can minimize the influence of shadow removal operation on other features in the shadow region. In shadow detection, the proposed method can be directly used for detection without training and ensures the consistency of similar regions to a certain extent. In this paper, each image is regarded as a small sample, and then a method based on material matching of intelligent computing between image regions is proposed to detect and remove image shadows. Shadow will cause some loss and interference to the information of moving objects, resulting in the performance degradation of subsequent computer vision tasks such as moving object detection or image segmentation. Shadow detection and removal play an important role in the field of computer vision and pattern recognition.
0 kommentar(er)
