New image-enhancing software for cameras allows scientists to more easily observe details under water. A team from the Netherlands Organization for Applied Scientific Research and the Royal Netherlands Navy used Nvidia Corp.’s CUDA (Compute Unified Device Architecture) graphics processing unit to develop new imaging software that uses algorithms based on imaging systems for compound security and unmanned aerial vehicles. Reference video footage of a harbor bottom. The original image is at top, while at bottom the same image is shown after contrast enhancement. Courtesy of SPIE. Traditional underwater cameras are typically hampered by low-contrast color changes, as well as interference due to camera noise and floating particles such as “sea snow.” This is especially true with cameras mounted on remotely operated vehicle systems, the researchers said. The new color correction algorithm developed by the team restores a scene’s original colors, however, each color correction must be designed from scratch for the specific underwater imaging situation. The software employs filters to reduce noise and correct contrast and color. The noise-reduction filter involves sub-pixel estimation of the camera’s motion, while adaptive image integration reduces any temporal noise. Small contrast variations are visible, but when amplified, this particular technique can achieve higher, evenly divided contrast of the entire image, the researchers said. The researchers collaborated with the Dutch Ministry of Defense to test the new approach in an operational naval environment, as well as against data from other similar underwater scenarios. The team is now working to further improve their new methods. They noted that because the noise reduction technique uses image content-reliant motion estimation, broadened optimization is possible if they employ the appropriate inertial measurement units. They have begun using a MEMS inertial measurement unit with active drift reduction. The researchers expect to deploy this technology by the end of 2015. The research was published in SPIE’s Defense and Security (doi: 10.1117/2.1201407.005484). For more information, visit www.tno.nl.