The vision system is important for almost all kinds of autonomous driving systems. However, visual data interfered by scattering media, such as smoke, haze, water, and other non-uniform media will be degraded seriously, showing the characteristics of detail loss, poor contrast, low visibility, or color distortion. These characteristics can significantly interfere with the reliability of autonomous driving systems. In real environments the image degradation mechanism is complex, and the estimation of degradation parameters is difficult. This issue remains to be solved. In this study, we employed dense blocks as the framework and introduced the attention mechanism to our model from four dimensions: Multi-scale Attention, Channel Attention, Structure Attention, and ROI (region of interest) Attention. With the help of the training data provided by the weakly supervised model, the proposed method achieved excellent performance in the task of scattering medium imaging optimization in different scenes. Comparative experiments show that the proposed method is robust, and is superior to other state-of-the-art methods in image dehazing, and underwater image enhancement tasks. It is of great significance to improve the reliability of autonomous driving systems in underwater and severe weather environments.
For underwater image restoration, modeling of PSF relies on a priori knowledge of the inherent optical properties under water. But full knowledge-based restoration could be corrupted by the complexity and instability of the application environments. Concerned with representation of the backscattering noise of underwater image, this Letter proposes a simplified multilayer transfer model to formulate both the point spread function and the statistics of backscattering noise explicitly. Consequently, the parameters for image restoration are estimated from in situ measurement of the backscattering background without a priori knowledge of the inherent optical properties. The robust Wiener filter is applied to implement the restoration. Experimental results are presented.
Structured lighting techniques have increasingly been employed in underwater imaging. Except for the limitation in underwater illumination, the desire of underwater image acquisition comes from the applicability of removal of the back scattered light, the critical problem of underwater imaging, using structured lighting techniques. This paper presents an approach for underwater image recovery using structured light and CCD camera. By integrating the scanned frame images, we generate an integration image which can be formulated as the convolution of the surface albedo and the illumination function. Thus underwater image acquisition is addressed as an optimization problem of image recovery and resolved by deconvolution, rather than the traditional geometric manipulation of frame tiling. The significance of the proposed method is that the forward scattering effect in the recovered image is fully eliminated by the integration operation and collection of the forward scattered light enhances the total imaging energy. By means of the structured lighting technique, an algorithm of using virtual aperture to limit the field of view is proposed to eliminated the back scattered light in frame images. Concerned with applicability of using the broad lighting pattern in structured light systems, the coded structured light with binary pseudo-random sequence is introduced, by which the high-frequency details in image recovery can be preserved through deconvolution. Results of underwater experiments are given.
20 February 2015 in Journal of Ocean University of China
This paper proposed an effective method called propagating deconvolution to recover single image degraded in a scattering medium. The propagating deconvolution is in the manner of the calculus, which is based on multi-layered decomposition of the scattering volume and modeled blurring function of a single-layer scattering volume. Parameters of the deconvolution algorithm are estimated just from in situ measurement of the pure scattered background from one single image.
2 May 2020 in Journal of Ocean University of China