A CBIR Scheme using new features of Color and Texture

Content Based Image Retrieval is a process of retrieving images from database using low level features. These low level features are color, texture, shape and spatial information. Two main features of the visual information are color and texture, since in most of the images relation between color and texture is very critical. There are several image retrieval schemes which employed many features with high feature vector dimension. However, they have achieved low precision. Here a new image retrieval scheme is proposed using color and texture features to achieve high precision by maintaining the low feature vector dimension. Color is represented by a new descriptor called Dominant Codebook (DC) and the texture is represented by Scan Pattern Co-occurrence Matrix (SPCM) and Scan Pattern Internal Pixel Difference (SPIPD). The DC describes the set of codewords and their percentages in the image. SPCM calculates the co-occurring probabilities of scan pattern between a pixel and its adjacent pixel in the image. SPIPD calculates the pixel difference with in the scan pattern. The color and texture features are combined to improve the retrieval performance.