The invention discloses a living example grade image searching method based on multi-layer characteristic representation. The method comprises the following steps: performing area extraction, namely dividing one image into multiple objects, and describing the whole image by using multiple objects, wherein the number of objects of most images is not greater than double digits at an object level; performing area screening; encoding characteristics; and performing index establishment. The invention discloses the concept of multi-layer characteristics of a depth structure, viewing from a biological angle, the mode that people observe objects in the nature is interpreted, the mode is extracted and applied to image searching modes, one image with multiple objects is divided into a plurality of areas, natural understanding of people upon image is met at the level of natural semantics, and user habits of people are relatively well met; and in addition, in a searching mode, the invention discloses a searching mode in an index mode based on KD (K-Dimensional) trees, and advantages and disadvantages of KDtree in indexing relative to those of lining indexing are deeply analyzed.