Low-level Features Extraction of an Image for CBIR: Techniques and Trends

Content-based Image Retrieval (CBIR) has gained much attention in the past decades. CBIR is a technique to retrieve images from an image database such that the retrieved images are semantically relevant to a query image provided by a user. It is based on representing images by using low-level visual features, which can be extracted from images such as color, texture and shape. Each of the features is represented using one or many feature descriptor. In this paper, Dominant colors in HSV and YCbCr and Color Coherent Vector (CCV) are implemented to describe the color feature of an image. For texture feature extraction, Gabor wavelet is best method. Texture features are found by calculating the mean and variation of the Gabor filtered image. Rotation normalization is realized by a circular shift of the feature elements so that all images have the same dominant direction. Another feature descriptor of texture is GLCM (gray level cooccurrence matrix). The features extracted from GLCM are its energy, entropy, contrast and inverse difference moment.