What Is a Depth Image (Depth Map)? How It's Made and Used

Left: the photo. Right: its depth map, made with our free Depth Map Generator (near is white). Photo: Marco Del Torchio 95, CC0, via Wikimedia Commons.

What a depth image shows

In Apple's words, a depth map is like an image, but "instead of each pixel providing a color, it indicates distance from the camera to that part of the image (either in absolute terms, or relative to other pixels in the depth map)" [2]. Related terms in 3D graphics are Z-buffer and Z-depth, the Z being the camera's line of view [1].

Each distance is shown as a shade of grey, but a depth map is not a greyscale photo. A maker on the Carbide 3D forum carving a horse from an ordinary picture (July 2021) found "the low spots I'd rather be the high spots"; the reply: "you need a depthmap (heightmap?), not simply a grayscale image" [17].

The number of shades matters. "An 8 bit depth map can only represent a range of up to 256 different distances" [1]; a 16-bit file has 65,536.

Make a depth map from your photo

The tool runs Depth Anything V2 Small in your browser, so the photo stays on your device. It saves near as white or black, in 8 or 16 bits; the full tool page explains the settings for engraving, CNC, Blender and ControlNet.

White is near, or black is near?

Both are in use. Wikipedia's example map is drawn with "Nearer surfaces are darker; further surfaces are lighter" [1]. The ControlNet depth model for image generation was trained on maps with "black representing deep areas and white representing shallow areas", that is, white near [3].

To check a map, find something you know is close. If it is dark and your software wants near as white (or the reverse), invert the map (in LightBurn, "Negative Image" [15]). An inside-out 3D effect, or an engraving that sinks what should stand out, means the same.

How is a depth image made?

Measured depth or estimated depth?

A sensor can give real distances: Apple's TrueDepth camera "can directly measure depth (in meters)" [2]. A model working from one photo usually gives relative depth: it predicts "the depth order of objects or points in a scene without providing the precise measurements" [11]; metric models exist too [4][11]. For blur, 3D effects, carving and image generation, relative depth is enough; to measure distances, you need a sensor or a metric model.

What are depth images used for?

• Portrait photos. A depth map lets software blur a picture more or less by distance [1]. Apple's example is the iPhone camera's Portrait mode; with the depth saved in the photo, the effect can be changed long after the shot [2].
• Unlocking your phone. Face ID builds a depth map of your face from its dot pattern [8].
• Robots, AR and games. Depth tells a robot where things are [6] and lets a console follow your body [9]; depth scanning also serves AR gaming, quality control and medical imaging [7].
• 3D graphics. Fog, shadows and Z-buffering use depth maps, and so do Magic Eye pictures (autostereograms) [1]; our Magic Eye Maker turns a depth map into one. With the camera's settings known, a depth map can also become a 3D point cloud [10].
• Turning 2D into 3D. "Most semiautomatic methods of stereo conversion use depth maps and depth-image-based rendering": from the picture and its depth map, software makes a left and a right image and fills in the areas that become uncovered [12]. Owl3D, the publisher of this guide, makes a desktop converter for Windows and Mac that turns 2D photos and videos into stereoscopic 3D with AI on your own computer, saving the 3D result, not a depth map; its free plan converts up to 60 seconds of video per conversion, with no watermark.
• Engraving, 3D models and AI images. LightBurn's 3D sliced engraving mode is "intended for use with depth maps" [15]; Blender's Displace modifier moves a surface by the intensity of an image [16]; ControlNet uses a depth map to guide image generation [3].

Why does my depth map look wrong?

Frequently asked questions

Is a depth image the same as a depth map?

Yes. Both name an image, or an image channel, that records the distance of surfaces from a viewpoint [1].

What is the difference between a depth map and a disparity map?

A disparity map records how far each point shifts between two side-by-side pictures [6]. Near things shift more [2], so disparity grows as distance shrinks. Apple's dual-camera depth data, for example, measures disparity in units of 1/meters [2].

Is a range image the same as a depth image?

Nearly. A range image's pixel values "correspond to the distance", and the sensor is "sometimes referred to as a range camera or depth camera" [13]. But some depth maps store distance to the camera's plane, not straight-line distance [1].

Does "image depth" mean bit depth?

Sometimes. Colour depth, "also known as bit depth", is the number of bits used for the colour of each pixel [14], a different thing.

Is a depth map the same as a height map?

In 3D graphics they are different tools: a depth map records distance from a viewpoint, while height maps and bump maps change a surface's shape or its look [1]. In engraving the two names meet: LightBurn's documentation speaks of "depth maps – sometimes called height maps" [15].

Can my phone take depth images?

Many can. On the iPhone, the dual camera, the front TrueDepth camera and the LiDAR camera record depth [2], and several other phones have time-of-flight cameras [9]. Any phone photo can also get a depth map afterwards from an AI tool such as our Depth Map Generator.