dResize

dResize [dRz] adjusts the spatial dimensions of a Deep Image. A Deep pixel may contain multiple Z-positioned samples, so this operation belongs in the Deep compositing pipeline rather than the ordinary 2D Resize branch.

Unlike an ordinary Resize that operates on a flat RGBA image, dResize changes frame dimensions while keeping the data in the Deep domain for depth-aware compositing downstream. It is not a creative image-transform or a one-click quality-enhancement tool.

The DaVinci Resolve 21 Deep Image chapter documents exactly one orange Image input on dResize. The previous article incorrectly added an Effect Mask input; the actual dResize diagram and tool signature should not contain that mask.

Controls contains two tool-specific fields: Width and Height. The bundled original Inspector capture displays 1920 and 1080, respectively. These are saved capture values, not a universal locked preset or a maximum supported resolution.

Width defines the output Deep grid in the horizontal direction; Height defines its vertical dimension. Keep the frame aspect intentional. A 960×540 → 1920×1080 adjustment preserves 16:9; merely changing 1920×1080 to 1080×1920 does not automatically reframe a horizontal composition.

dResize and dTransform are different. dTransform can move Deep imagery and adjust depth using Scale Z and Translate Z. The dResize Inspector does not document Z Offset, Pivot, Size X/Y, an interpolation selector, or a mask input.

A Deep OpenEXR can contain multiple sample intervals per pixel. After changing width or height, evaluate thin translucent edges, overlapping objects and fine coverage details. The manual does not specify a particular Deep resampling filter, so the article must not invent one.

A typical scenario is matching a half-resolution Deep smoke pass with a 1920×1080 compositing target. Set dResize to the expected dimensions, but also check frame timing, pixel aspect, camera position and Z conventions before feeding the stream into dMerge.

Matching frame size alone cannot correct different projections, incorrect Z scaling, a shifted camera, mismatched animation frames, or incompatible color handling. Check these independently when an overlap is wrong.

Deep Image nodes are part of the DaVinci Resolve Studio toolset. A real Deep EXR can arrive from a suitable Loader/MediaIn or rendering route. For a teaching example, Background → Image to Deep offers explicit 2D-to-Deep conversion, but does not reconstruct the hidden surfaces of a true multi-sample EXR.

The output of dResize remains Deep. Keep sending it to suitable Deep processing (for example dMerge or dRecolor), or use DeepToImage when a conventional RGBA image is required for a 2D Merge or MediaOut.

An A/B method uses two branches of the same Deep source: an original source → DeepToImage reference and a source → dResize → DeepToImage comparison. Compare at the same timeline frame and 100% Viewer zoom. Flat output checks are useful, but cannot prove that every Z sample survived exactly as intended.

Increasing the width and height can be expensive. Double each dimension and the number of pixel locations grows fourfold, although actual Deep-memory usage also depends on the non-uniform count of depth samples per pixel.

A lower-resolution Deep proxy can help iteration but resizing the proxy back up is not a lossless reconstruction of a high-resolution original. Return to the original source for final-quality work.

The shared Deep Image Settings controls include Hide Incoming Connections, Comments, and Scripting. Only the Controls tab is visible in the bundled dResize screenshot; exact Settings layout for the installed application build is not asserted from that image.

The four lessons below are static, typed graph exercises. They do not claim completed Resolve GPU rendering or real Deep EXR occlusion tests. Live verification must use an actual multi-sample EXR in the target Resolve Studio version.