Feature extraction is a sort of dimensionality reduction that efficiently represents region of an image as a compact feature vector or the process in which certain features of interest within an image are detected and represented for further processing. This research work proposes a hybrid feature descriptor based on color, texture, and shape. The feature set includes two color features: Average Color Difference (ACD) & Color and Edge Directivity Descriptor (CEDD), a texture feature using Local Binary Pattern (LBP) and shape feature using Zernike Moments (ZM). Image descriptors derived from different color spaces often exhibit different properties, among which are high discriminative power and relative stability over the changes in photographic conditions such as varying illumination, hence, the color features are derived from different color spaces like YIQ, HSV and L*a*b. Then the feature vectors are normalized and fused to improve the classification performance.