In content Based Image Retrieval, images are retrieved based on color, texture and shape (low level perception). There is a gap between user semantics (high level perception) and low level perception. Relevance feedback (RF) learns association between high level semantics and low level features. Bayesian method, nearest neighbor search method, Log based RF, Support Vector Machine (SVM) is methods of Relevance Feedback. Bayesian method is good for understand but it is not worked for fast access. In nearest neighbor search method, data are compressed. So some times images may be lost its features. Log based method are used soft label using SVM. SVM is best method for RF because it works on structure risk minimization by VC dimensions.