De-noising of near infrared image and nonlinear estimation of nitrogen content were carried out to cucumber leaves in greenhouse. Fruit cucumber Deitastar was chosen as the object. A CCD camera with special filters was used to collect cucumber leaves’ images in different growth time. After eliminating noise of image with wavelet transform, the images were separated based on grey theory. Correlation analysis between nitrogen content and each kind of vegetation index of cucumbers was conducted, and t tests to those coefficients of correlation were executed. The result showed that CNDVI, GNDVI, NDVI and NDGI were significantly related to nitrogen content of cucumber leaves. The correlation coefficient between CNDVI and nitrogen content of cucumber leaves reached 0.67, and the correlation coefficients between GNDVI, NDGI, NDVI and nitrogen content of cucumber leaves were all higher than 0.50. LS-SVM algorithm was used to construct estimation models between vegetation indexes and nitrogen content and CNDVI, GNDVI, NDGI and NDVI were used as the input of the model. The R2 values of calibration and validation models were 0.825 and 0.728 respectively.