A computerised thermal imaging system for thyroid diagnosis was developed by the authors and it was discovered that the rate of changes of temperature, rather than the absolute values, associated with a sequence of thermograms could help the medical doctors to identify clinical disorders. In order to further enhance the diagnostic capabilities and speed, a new method for medical thermogram analysis has been developed that compresses a sequence of thermograms into one thermogram while retaining the important information such as the geometrical patterns of the objects and the rate of temperature changes of each pixel within the images. As motion artifacts are unavoidable when a patient undergoes minutes of thermogram recording, direct comparison between images is deemed impossible. A high speed image matching algorithm has been developed to provide an absolute geometrical foundation for pixel-to-pixel comparison. The rate of change of temperature of a particular pixel along the sequence is represented by one single parameter after a process of temperature integration which can then be converted into a corresponding gray level for display. The resultant compressed thermogram can give a clear distinction between problem areas and normal ones. Although our emphasis is on thyroid diagnosis, it is anticipated that this new technique can be applicable to other areas of a human body.