The underwater world is one of the most fascinating and least explored. As we plunge into the blue waters for explorations, the best of the imaging cameras and lights often fail to capture the best on the polaroid. It’s not the fault of cameras or lights but it is the very medium – water and the suspended particles which play havoc with the photography or videography. Underwater images suffer from lack of proper visibility due to wavelength-dependent light attenuation and scattering. Another factor is the type of water condition because of turbidity, which makes the underwater image/video data too difficult to process. The current enhancement techniques try to obtain better visibility of features in the image without affecting the overall quality of images.
Beyond recreational underwater photography, imaging has many applications in a variety of fields like marine research, underwater inspections of structures, ships, underwater robotics-based mapping of seafloor etc… We need clear underwater images/video for identifying different types of structures and aquatic life to do any useful assessments. Underwater robotic systems also rely heavily on high-quality images to fulfil their mission objectives. Many factors degrade the quality of the acquired images One of the major factors for the degradation of images is wavelength-dependent light attenuation along with the depth of the object in the scene. For example, red light is absorbed in the water at a higher rate than blue or green light. Hence, we see a blueish or a greenish tint in an underwater scene. Another factor diminishing underwater image quality is the light scattered due to the small particles present in water, which introduces a homogeneous background noise/haze to the image
The other problem is about the type of water in which we’re taking the image, Which has different characteristics. You can see how the images differ from each other from deep-oceanic environments to shallow muddy environments
There are two different enhancement approach to solve these
Using the image processing techniques like colour-correction, Equalization etc
Deep Learning Methods
Existing image processing techniques
These methods modify the data according to the specific set of instructions based on different algorithms/methods which will have to be selected based on different conditions and after analysing the image characteristics. If a method shows good results in a particular type of water, the same method will not work on another type of water. So while using the above image processing techniques or the methods can solve a particular problem in underwater image enhancement and will work only on the similar or same type of water for which it was developed. This is a major drawback of these techniques.
Deep-Learning techniques
These methods try to generalize the enhancement of underwater images and are not programmed for a specific condition, These techniques use deep-learning techniques like CNN, GAN for enhancement. While using these we will try to give the different image data to the network and the network will learn about using the corresponding ground truth. So the advantage is, we will get a model which will be good at different types of water But the key drawback or shortcoming is that requirement of large datasets for training the model, as it is not that common. As the deep learning methods require good dataset, we lack the availability of such underwater image dataset. To get around this, at EyeROV we have created training data on a synthesised dataset. We have generated underwater environments and conditions on normal images, thus creating a pool of images in varying water conditions. Deep learning techniques with these methods need to mature further to be effective.
EVAP
EyeROV’s visualization and analytics platform – has one major component as underwater images/videos enhancement without compromising in the quality of data. We’re using both the image processing technique and deep learning methods for getting the best of both worlds. EVAP also has other functionalities like video overlaying useful user data into the videos for better understanding and quick interpretation of data.
With the rapid advancements in imaging techniques and specialised photography hardware and tools, the future seems more crisp and clear for the underwater world. We at EyeROV thrive to stay on top of the wave to catch hold of the advanced techniques and deliver the best results to our clients.
