In this study, we propose a new regular-texture image retrieval approach by which users can retrieve regular-texture images from a database which are most similar to a sample query. In this approach, texture primitives and their displacement vectors are extracted from the query and each regular-texture image in the database. These components describe periodic properties of a regular-texture image. Five features are computed from co-occurrence matrices (CMs) of the texture primitive to characterize statistical properties of the corresponding image. Each regular-texture image in the database is then represented as the five CM-features, which are insensitive to translation and rotation of the regular-texture image. Hence, query comparison or matching can be done using the corresponding CM-features. Experimental results show that the proposed approach is indeed effective. The time required to process a query is moderate.