A pairwise image analysis with sparse decomposition

摘要: This paper aims to detect the evolution between two images representing same scene. The detection problem has many practical applications, especially in medical images. Indeed, concept of a patient “file” implies joint analysis different acquisitions taken at times, and detection significant modifications. The research presented this is carried out within application context development computer assisted diagnosis (CAD) applied mammograms. It performed on already registered pair As registration is never perfect, we must develop comparison method sufficiently adapted real small differences between comparable tissues. In assessment similarity used during step also for the interpretation that yields prompt suspicious regions. our case assumed match spatial coordinates similar anatomical elements. paper, order process tissue level, image representation based elementary patterns, therefore seeking not pixels. Besides, as studied have low entropy, decomposed signal expressed parsimonious way. Parsimonious representations are known help extract structures signal, generate compact version data. change should allow us compare short time, thanks weight thus represented, while maintaining good representativeness. precision results show approach efficiency.