In biometric research and many security areas, it is very difficult task to match the images which is captured by different devices. Large gap exist between them because they relates with different classes. Matching optical face images to infrared face images is one of the difficult task in face biometric. Large difference exists between infrared and optical face images because they belong to multiple classes. Converting the samples of multimodality into common feature space is the main objective of this project. Different class of images is relating by coordinating separate feature for classes .It is mainly used in heterogeneous face recognition. The new method has been developing for identification of heterogeneous face identification. Training set contains the images from different modalities. Initially the infrared image is preprocessed by applying Gaussian filter, difference of Gaussian and CSDN filters are apply on infrared face image. After preprocessing next step to extracting the feature by using LBP(local binary pattern) feature extraction then relevance machine classifier is used to identify the best matching optical image from the corresponding infrared images from the optical images dataset. By processing this technique our system efficiently match the infrared and optical face images.