Abstract: Eestimating their weight and size, therefore valuing their growing steps, based on snapped their photographs, is an efficient and quick method in order to meet the information requirment for process controlling in industrial aquaculture. For this purpose, effective restoring methods to process underwater degraded images are essential for this technology. Based on polarization imaging technology and improved turbulence model, a robust model for underwater degraded images restoration was proposed in this paper. Firstly, aimed at increasing the completeness of prior knowledge required by restoring degraded images method, an improved underwater turbulence model was designed by considering the wave structure function and distribution function of scattering scale parameter to overcome the shortcoming resulting from the simplified turbulence model, which simply imitated the situation for atmospheric turbulence and graded turbulence intensity only depending on fuzzy factor. Secondly, noise characteristics in degraded images was drawn from comparison between “and “ , “minus” images, which were computed from polarization images, based on polarization characteristics of underwater forward scattering light in visible band. For the precision of noise characteristics drawing, pulses coupled neural network (PCNN) and wavelet transfer (WT) algorithm was applied in computing the “and “ and “minus” images. And then, the algorithm, namely constrained least squares filtering (CLSF) method, was applied to replace Wiener filtering for restoring the degraded images because of its robust characteristics. Finally, comparison of restoration results among the four different restoring methods was carried out to evaluate the effect of our proposed method according to four valuating parameters, namely average value (AVG), standard deviation (SD), Entropy and signal-to-noise ratio(SNR). The experimental conditions, especially for turbulence situations, were designed as: feeding Chinese carp with weight of about 0.5 kg at a pool (5 m×4 m×2 m). All images were snapped at 30 cm depth under water surface. The results showed that more ideal restoration effect in strong turbulence circumstances could be realized based on the proposed method, i.e. improved turbulance model accompanied with CLSF This should be beneficial for further research works on underwater degraded images restoration in complex flow conditions.