Analyse de la texture des images mammaires par une fusion des lois de Zipf et des Ondelettes pour la classification des tumeurs mammaires via l’analyse en composantes principale

Abstract

ان اختيار العلاقات عدم الخطية له أهمية كبيرة بالغة في تطوير ادوات قوية التحليل الصور والرؤية عبر الكمبيوتر مشكلتنا البحثية هي تطبيق القوة اف زيت وزيت اف معكوس لتحليل صور الثدي وتتميز القوة التعقيد الهيكلي لبنية الصورة عبر نمدجة الاحصائيThe choice of the non-linearity is of crucial interest in the development of powerful tools for image analysis and computer vision. In this sense, our research problematic is to apply the power laws: Zipf and inverse Zipf for mammogram images analysis. Indeed, the laws of Zipf characterize the structural complexity of the image texture by modeling the statistical distribution of patterns frequency of appearance as power law distribution. In addition, we have performed a fusion of the obtained texture features with those generated once applying Haar wavelet transform for mammogram images analysis .We will apply a PCA principal component analysis to reduce the number of descriptors. Subsequently, we have proposed a content based mammogram image indexing and retrieval system (CBMIIR) that boosts the performance of a computeraided diagnosis (CADx) at the stage of providing the diagnostic to radiologists. Indeed, radiologists feel more confident in their diagnosis decision based upon case-adaptive classification via the template-matching technique, where similar known cases, to the one under analysis, are retrieved and displayed from indexed databases; rather than the abstract result generated by a classifier. The evaluation of the proposed systems has given encouraging performance