The objective of this research was to study the potential of NIR diffuse reflectance spectroscopy for nondestructive measurement of chlorophyll content in pepper leaves. Statistical models were developed using partial least squares (PLS) and principal component regression (PCR) analysis technique, and four different mathematical treatments were used in spectra processing: smoothing, baseline correction, different wavelength range, and first and second derivative. The results showed that smoothing and baseline correction of NIR original spectra had little influence on spectral data analyzing, second derivative spectra showed better results than first derivative spectra and original spectra, while PLS analysis could get better modeling results than PCR analysis. The best calibration model was obtained after removing four outlier samples, and processing spectra with second derivate in the full wavelength range, with the highest correlation coefficient (r = 0.97537), a relative lower RMSEC value (2.33), and a small difference between RMSEC (2.33) and RMSEP (5.49). It can be concluded that NIR technique is a feasible, nondestructive way to predict the chlorophyll content in pepper leaves.