Egg freshness is an important index to measure the economic quality of eggs, and 98 eggs’ image information of different storage time were collected. In the meantime, the conventional index, includes the weight, the height of protein and chamber of the egg which reflects the egg freshness were measured. To receive a method of nondestructive examination, using MATLAB image processing toolbox as a research tool to measure the area ratio of egg yolk, chamber height and length of the whole egg size. Analyzing the regression model between weight and chamber height, whole egg’s major-minor axis, the regression model between protein height and yolk area ratio, chamber height, etc. by SPSS; The egg weight and albumen height were selected to be characteristic parameter to establish the prediction model of fresh grade with DBN(Deep Belief Network), and the well trained network is used to estimate the degree of freshness of the corresponding target. Experimentally found that the correlation coefficients were 0.942 and 0.925 which made by the prediction model of weight and Albumen Height, and the DBN’s classification accuracy of fresh-ness grade is 93.3%. The results showed that the design of egg freshness prediction based on image processing data has high accuracy and low cost, which can be used to estimate egg freshness.