This study aims to use various types of color invariants as features to implement image retrieval. The invariant features are used to build normalized 3-D color histograms as image indices for the query image and each of the reference images in the database. Then, matching is done using normalized histogram intersection. The performances of various color invariants are compared. Experiments indicate that the method effectively distinguishes identical objects that are rotated, illuminated from different directions, or viewed from different angles from all other objects. It was also discovered that the features are more resilient to variations in both illumination and orientation when recognizing non-shiny objects compared to arbitrary and shiny objects. Finally, we observed that the invariants are not effective at finding similar objects.