Abstract: In order to solve the problem of estimating the weight of the flounder and other flatfishes underwater, the image and weight data of the tonguefish (Cynoglossus semilaevis) were obtained at different growth stages. The area of the fish were measured through the image processing. The fitting models were established to match the area comparaed with weight. The results showed that the correlation between the area and the weight could reach at 0.9682, the average relative error was 6.17%. In addition, the weight was also affected by other shape parameters, and the ration of equivalent ellipse axes and heywood circularity factor were measured. Three-dimensional fitting models of area, ration of equivalent ellipse axes, weight, and area, heywood circularity factor, weight were established. The average relative errors of the two models were 5.50%, 5.62%, respectively. The experiment verified that the processed results of images of fish underwater obtained out of the surface and calibrated with the template underwater were consistent with that of images obtained outside the water. Consequently, the weight of flounder fish underwater can be estimated without catching the fish.