Image texture is often characterized using gray level co-occurrence matrices (GLCM). The GLCM statistics reflect only the highest power and spatial frequencies. To address this, researchers have employed discrete wavelet transform (DWT) along with GLCM. However, this method involves a computationally complex convolution operation in the spatial domain, and also inherits the sampling limitations of the DWT. Extending texture analysis to the space-frequency (SF) domain will uncover patterns not visible through the GLCM-based approaches while still capitalizing on the effectiveness of the traditional co-occurrence matrix. The discrete S-transform (DST) provides the SF representation at a pixel by localizing with a Gaussian modulated sinusoidal window. The DST based texture analysis is proposed to improve upon the GLCM while providing advantages over wavelets. This paper presents the promising preliminary results achieved using the proposed method.