Design And Research Of Automatic Image Enhancement Of Deep-sea Environment

In recent years, with increasing interest of scientists in the exploration of ocean, Submarine detection technology developed rapidly. Underwater operation machine collected video information and transmitted it to the host computer, the video plays an important role in the monitoring of marine life, geological assessments, environmental monitoring, and biological resources of the seabed exploration and so on.The light source of underwater imaging device in shallow sea has two aspects: one is the light reflected from the target surface directly transmitted to the camera, the other is the background light which is formed from multiple scattered by the suspended particles. The suspended particles are composed of sand, mineral particles, plankton and others components, these suspended particles dispersed light beam that reaches the camera, resulting in a blurry image quality. Another problem of the shallow imaging is image color change. The light attenuation in the process of continuous submarine transmission, attenuation amplitude increases with the increase of depth. While the longer-wavelength decay faster, the shortest wavelength of the blue light, so the slowest attenuation, therefore the undersea video images appears more blue. In addition to these problems, the existence of overexposure and underexposure phenomenon is another problem of the undersea video image. The reason is that in the deep dark environment, natural light attenuation is exhausted already, therefore we must use artificial illumination to provide light for imaging equipment, which resulted in the image appears the problem of uneven illumination. Deep-sea video image has image blur, color variation, and overexposure and underexposure phenomenon caused by uneven illumination simultaneously. All the time, the image processing is committed to onshore processing levels, this topic mainly aims at these undersea characteristics for algorithm research, on the basis of ensuring the image quality, and we design an image enhancement algorithm.The main work of this paper is, firstly decompose the input video into images, then using the no-reference perceptual quality assessment of JPEG compressed images algorithm to assess the image quality. If the image through the quality evaluation, illustrating that it meets the need of following research which can be output directly, otherwise, it must be enhanced by follow-up method in this paper.Secondly, after the image quality evaluation, we use the corresponding algorithm to eliminate the image blur. Because of lack of underwater light, haze is one of the common problems of underwater images. In this paper, we use single image haze removal based on dark channel prior algorithm to dehazing. After this step, image haze is eliminated basically, however, Due to the uneven illumination phenomenon during the seafloor imaging, overexposure and underexposure is widely exist in undersea images, so we still need to take algorithm for image enhancement.Lastly, on account of over/under-exposed regions exist in undersea image widely, in this paper aims at this problem, an image enhancement algorithm is proposed. The algorithm firstly extracted the over/under-exposed regions of undersea images, and then calculated the upper and lower limit of the two regions respectively. Secondly, calculating the image dynamic range to determine whether the image exist “idle dynamic space”. Finally, for those images who exists “idle dynamic space”, we precede image dynamic range expansion, while increasing the dynamic range of over/under-exposed regions, next the image was converted from RGB to YCbCr color mode to separate Y component for Histogram equalization. Then fused the three channels of image, converted the image to RGB color mode, display the output image.After programming and debugging, we draw the conclusion that compared to the traditional enhancement algorithm, the new method not only avoids unnatural results of image caused by traditional histogram equalization, but repairs missing pixel information, improves the clarity and details of images significantly, besides, the algorithm has a strong targeted on the undersea images.