This work describes the development of a flame analysis and characterization technique based on the application of invariant transform to the measured optical data of flame images. In this technique, we utilize matrix-norm analysis to obtain the scalar quantity of the post-transform images, which are specific to each global equivalence ratio, or flame state. To demonstrate the potential of the proposed technique, the invariant transform of a time-resolved projection image of a CH4/air flame captured by a CCD camera was compared with those obtained by computed tomographic reconstruction (CT) at three different equivalence ratios: 0.97, 1.09, and 1.29. The results show that the invariant transform can determine flame states more effectively than the projection and reconstruction data. It is also found that if care is taken in selecting an appropriate nonlinear kernel function for a particular flame image data, a tradeoff between computational time and the ability to extract flame information is optimized. Thus, it is feasible to implement the technique as an online visualization and analysis tool for combustion process monitoring and control.