We present an approach for human face recognition using eye region extraction/replacement method under low illumination and varying expression conditions. For conducting experiments, two different sets of face images, namely visual and corresponding thermal, are used from Imaging, Robotics, and Intelligent Systems (IRIS) thermal/visual face data. A decomposition and reconstruction technique of Daubechies wavelet co-efficient (db4) is used to generate the fused image by replacing the eye region in the visual image with the same region from the corresponding thermal image. After that, independent component analysis over the natural logarithm domain (Log-ICA) is used for feature extraction/dimensionality reduction, and finally, a classifier is used to classify the fused face images. Two different image sets, i.e., training and test image sets, are mainly prepared using the IRIS thermal/visual face database for finding the accuracy of the proposed system. Experimental results show the proposed method is more efficient than other image fusion techniques which have used region extraction techniques for dark faces.