Content based image retrieval (CBIR) has been the center of interest for a long time. A lot of research have been done to enhance the performance of these systems. Most of the proposed works focused on improving the image representation(bag-of-features) and classification methods. In this paper, we focus on enhancing the second component of CBIR system: region appearance description method. In this context, we propose a new descriptor describing the spatial frequency property of some perceptual features in the image. This descriptor has the advantage of being lower dimension vs. traditional descriptors as SIFT (60 vs 128), thus computationally more efficient, with only 5% loss in performance using a typical CBIR algorithm on VOC 2007 dataset. The number of digital images continues to increase, especially with the expansion of social networks: according to Time magazine, more than 130,000 images are uploading each minute on Facebook. Thus, it will be difficult for a human to use this vast collection of images, e.g.: searching manually for images containing objects or persons. Content based image retrieval(CBIR) system is necessary for this kind of tasks. Content based image retrieval (CBIR) has been the subject of interest in the computer vision community for a long time: a lot of algorithms have been proposed in the last decades. Most of these systems are based on local approaches (illustrated in figure 1). According to [6], local approaches consists in 5 steps: region selection, region appearance description, region appearance encoding, derivation of image features from the set of region appearance codes by spatial pooling, classification. A baseline of CBIR method [6] is represented in figure 1. In the following, we present the different methods used in each step. region selection As shown in figure 1, a typical CBIR algorithm first scans the image to select the regions of interest. In the literature, there are two concepts to accomplish such task: