Object detection in images is a fundamental task in many image analysis applications. Existing methods for low-level object detection always perform the color-similarity analyses in the 2D image space. However, the crowded edges of different objects make the detection complex and error-prone. This paper proposes to detect objects in a new edge color distribution space (ECDS) rather than in the image space. In the 3D ECDS, the edges of different objects are segregated and the spatial relation of a same object is kept as well, which make the object detection easier and less error-prone. Since uniform-color objects and textured objects have different distribution characteristics in ECDS, this paper gives a 3D edge-tracking algorithm for the former and a cuboid-growing algorithm for the latter. The detection results are correct and noise-free, so they are suitable for the high-level object detection. The experimental results on a synthetic image and a real-life image are included.