Molecular spectra analysis combined with the chemometrics is becoming a popular method for rapid classification of edible oil. However, when the molecular spectral differences among the different types of samples are tiny, it is usually difficult to identify them with the traditional classification techniques. In this work, a method of molecular spectra analysis based on image recognition for rapid classification of edible oil is proposed. In order to accomplish recognition of different types of edible oil, the attenuated total reflectance infrared spectra of seven types of edible oil are scanned on ATR-FTIR. To enhance the spectral differences among different types of samples and visualize the identification process, the pretreated IR spectra are transformed into two-dimensional spectral image with auto correlation operation. Then, the local extrema are extracted with the method of image expansion and are used as the classification features. The back propagation (BP) neural network is chosen as the classifier to identify the extracted local extrema of the two-dimensional spectral image. Comparative experiments to identify the same samples with the proposed method, PCA-BP and KL-BP have also been done. Comparative experiment results have verified that the classification results with the proposed method (correct classification rate is 94.4%) are obviously better than those with PCA-BP (correct classification rate is 66.7%) and with KL-BP (correct classification rate is 83.3%). The proposed method has provided a new way to classify the edible oil rapidly based on molecular spectra analysis.