We present an image coding method explicitly designed for easy content access, i.e., content-based image retrieval. Based on a number of well-studied conventional image coding methods, namely segmentation-based image coding (SBIC), vector quantization (VQ), and a recently developed coloured pattern appearance model (CPAM), we have developed an image coding method with a compressed stream from which effective image content descriptors can be derived with very little computation. A colour image is first segmented adaptively into homogeneous regions of various sizes. Each region is then decomposed into three channels according to the CPAM and VQ is employed to represent the chromatic and achromatic spatial patterns efficiently. The image content descriptors are the joint probability distributions of the segmented region sizes and their achromatic and chromatic spatial patterns’ VQ codebook indices. From image indexing and content-based retrieval perspective, this work can be regarded as a method effectively exploiting/employing image coding technologies to develop novel and effective image descriptors for content-based image retrieval. We have applied the newly developed image content descriptor to content-based image retrieval from a large colour photo image database. Experimental results demonstrate that the new method is comparable to state of the art methods, such as colour correlogram and the latest MPEG7 colour structure descriptor in content-based image retrieval.