In this article, an algorithm for a health diagnosis system for the elderly that uses image processing was developed. This algorithm reconstructs monochrome images from data of the time a subject watches TV and calculates the index for diagnosis of the subject's health condition from the entropy of the image. When this algorithm was applied to the data obtained for seven months, the result almost corresponded to the health condition of the subject. It is assumed that this method can be used not only for diagnosis of a physical condition but also for diagnosis of a mental condition. As a result of the evaluation, this algorithm is considered to be effective according to the subject. In addition, this method was economical because the algorithm needed only simple data acquired from simple sensors. In the future, an automatic diagnosis will be available using this algorithm. Finally, total medical expenses will be reduced if this system is practical.