Query by image contents is a successful approach to image database indexing and retrieval. Because image data take up huge amount of storage space and network bandwidth, compressing image is a standard practice. Given that images are compressed/coded in a database, it would be more efficient if the compressed data could also be directly used as indices for content-based query. The compressed data of traditional image coding/compression techniques such as the popular JPEG cannot be readily used to classify image contents. In this work, we set out to develop image coding schemes that will enable classification of image contents to be performed in the compressed domain without the need to perform image decoding and any extra processing. One of the novel features of this work is that we exploit some of the well known properties of human colour visual system for image representation. We outline a general scheme of colour image coding for classification and describe a specific implementation.