A face recognition algorithm based on a newly developed Transform Domain Mutual Principal Component Analysis (TD-2D-MuPCA) approach is proposed. In this approach, the spatial facial two-dimensional images (2D) and their division into horizontal, vertical and diagonal sub-images halves are generated. The sub-image halves are processed using non-overlapping and overlapping windows. Each face and its processed sub-images are subsequently transformed using a compressing transform such as the two dimensional discrete cosine transform. This produces the TD-2D-MuPCA. The performance of this approach for facial image recognition is compared with the state of the art successful techniques. The test results, for noise free and noisy images, yield recognition accuracy of 97% or higher. The improved recognition accuracy is achieved while retaining notable savings in storage and computational requirements.