Underwater images suffer from low visibility and contrast caused by absorption and scattering, which leads to haze and some further limitations. The existing underwater single image dehazing methods cannot achieve a balance between the performance and computational complexity, and are difficult to produce satisfactory results in the regions with large distance. To overcome these problems, we propose a new underwater single image dehazing method, which includes an improved background light estimation based on the quad-tree subdivision iteration algorithm, and a novel transmission estimation method. For the background light estimation, we introduce a robust score for each region of the image, which can evaluate the region from both smoothness and color. For the transmission estimation, we propose the color space dimensionality reduction prior (CSDRP), which allows conversing an image from the three-dimensional RGB color space to a 2D color space, namely the UV color space. In the UV color space, by clustering the pixels into mounts of haze-lines and carefully setting the haze-free boundary, the transmission map can be figured out and used to produce an excellent dehazed image. Experimental results show that our method has competitive effects compared with mainstream underwater single image dehazing methods. INDEX TERMS Underwater image dehazing, contrast enhancement, image enhancement, scattering removal.

Absorption and scattering in aqueous media would attenuate light and make imaging difficult. Therefore, an artificial light source is usually utilized to assist imaging in the deep ocean. However, the artificial light source typically alters the light conditions to a large extent, resulting in the non-uniform illumination of images. To solve this problem, we propose a non-uniform illumination correction algorithm based on a fully convolutional network for underwater images. The proposed algorithm model the original image as the addition of the ideal image and a non-uniform light layer. We replace the traditional pooling layer with dilated convolution to expand the receptive field and achieve higher accuracy in non-uniform illumination recognition. To improve the perception ability of the network effectively, the original image and parameters which pre-trained on the ImageNet are concentrated. The concentrated information is used as input to the network. Due to the color shift and blurred details of the underwater image, we design the novel loss function, which includes three parts of feature loss, smooth loss, and adversarial loss. Moreover, we built a dataset of the underwater image with non-uniform illumination. Experiments show that our method performs better in subjective assessment and objective assessment than some traditional methods.

Abstract-This paper presents the factor clustering analysis for violent crimes. The efficiency of Rough-fuzzy C-means algorithm is affected by the numbers of clusters, and not all centroids are beneficial. The analyzing of violent crime data does not need human intervention for impartiality. The information entropy is a helpful tool for resolving those issues. In this paper, a novel discrete Rough-fuzzy C-means based on information entropy algorithm (DRFCMI) is proposed, which can obtain typical conclusions objectively. Experimental results illustrate that our proposed method is efficient.

No abstract

No abstract

No abstract