| Deep sea exploration and operation technology is one of the key research fields for ocean research and development. The commonly used underwater observation methods include acoustic sonar based long distance observation method and optical video based short distance observation method. The former can operate in larger visibility range at expense of a coarse resolution, thus mainly suitable for underwater objects searching in a large area of waters. The later provides images with high resolution, rich texture and color information, thereby being in favor of object recognition and understanding. Therefore, optical video imaging is one of the key techniques for underwater short distance inspection and operation. However, underwater images have two problems. One is that underwater images exhibit poor contrast and blurry details due to the scattering and attenuation effect of the light underwater. The other is that underwater images are with non-uniform illumination, such as bright specks and dark regions in the image, due to the artificial lighting and the movement of underwater platforms. These problems decrease the visibility of the underwater images, as well as bring challenges for the subsequent image auto processing and recognition processes. Hence, the thesis has studied the visibility improving methods of underwater video images, and has finished the following research work.(1) Research of underwater image restoration algorithm based on physical imaging model. According to the underwater physical imaging model, a visibility improving algorithm for underwater degraded images is proposed based on wavelet transform. Firstly, the RGB color space is transformed to the YUV color space. Then, for the luminance component Y, according to the contrast of the original image, the amount of media scattering light is estimated adaptively by using wavelet transform to enhance the contrast of degraded underwater images. Moreover, non-linear luminance adjustment is implemented on the low-pass subband of wavelet transform to eliminate the problem of non-uniform luminance. Finally, the visibility enhanced underwater image is obtained by combination the resulted image of Y component and the chrominance components of U and V. The experimental results show that underwater degraded images can be restored adaptively by the proposed algorithm.(2) Research of underwater image enhancement algorithm based on light reflection model and characteristic of image neighborhood. In order to solving the non-uniform lighting problem of underwater images, an image enhancement algorithm is developed based on light reflection model and characteristic of image neighborhood. Firstly, for the illumination component image Y, the component of irradiation part is estimated by using low-pass subband of the wavelet transform of the Y image, and the original non-uniform Y image is compensated from the estimated result. Then, the contrast of underwater degraded image is improved based on the characteristic of the image neighborhood. The experimental results show that the proposed algorithm can both increase the contrast of the degraded underwater image and effectively overcome the non-uniform illumination problem.(3) The transplantation and optimization of the visibility improving algorithm to the DSP based hardware system. To meet the practical requirement of DSP based embedding underwater systems, the driver program of the TMS320DM642 video port is developed based on the characteristics of TMS320DM642 DSP video port peripheral. The underwater video restoration algorithm developed on the VC++6.0 platform is transplanted to DSP embedded platform, and various DSP optimization techniques are comprehensively used, such as compiler options, intrinsics, loop unroll, and so on. Finally, based on DSP/BIOS real-time operating system, the real-time performance of the developed algorithm is tested, and several improving measures are proposed for the future improvement of the algorithm. |
PDF Full Text Request