Full frame texture images alongside a cropped 3D scan

A cropped depth result does not tell you whether the camera retained a full-size texture exposure. The crop may happen at the sensor, during rectification, or only during depth processing. Each location changes what remains available for a separate inspection.

For an integration that needs both depth and conventional image analysis, establish image availability before changing calibration or adding another camera. The workflow below identifies retained texture data, checks whether it can be processed without another exposure, and defines when a separate acquisition is necessary.

Separate the acquisition, rectification and depth regions

An image called “texture” describes its intended use, not necessarily its acquisition format. It might be raw or rectified, monochrome or color, full-size or cropped. A texture exposure may also occur at a different time from the projected-pattern exposures used for depth.

Distinguish these regions before troubleshooting:

These operations can be coupled by the acquisition software. A setting labeled “scan ROI” may change several stages together rather than controlling depth alone.

The related ROI, binning and subsampling comparison explains why reduced output dimensions do not identify the underlying acquisition method.

Before testing, gather the documented stream descriptions, image metadata, current configuration and a stationary target. Use the manufacturer's supported tools and documentation to establish which raw outputs are actually exposed.

Record a full-size baseline

1. Capture a baseline without the scan crop, using the supported texture acquisition workflow.

Save the original image buffers rather than screenshots or viewer exports that may already be resized. Record the dimensions, pixel format, row stride, timestamps and frame identifiers available for each output. Include the camera or sensor identity when several imagers contribute to one scan.

Build a small acquisition record containing:

• Raw texture images, if the interface exposes them.
• Rectified texture images and their coordinate frame.
• Pattern images or other depth inputs, where accessible.
• The depth output and its processing region.
• Calibration identity and the effective acquisition settings.

Do not infer full-size coverage from buffer length alone. Row padding, packed pixels and metadata can change the byte count without changing image coverage.

Place visible landmarks near the top, bottom and center of a stationary target. Their presence helps distinguish a true full-frame image from a resized crop. For stream identification and metadata checks, use the raw stereo capture verification guide.

Locate where the crop takes effect

2. Apply a clearly smaller processing region, repeat the capture, and compare every available image with the baseline.

For example, retain the central third of the image height while leaving the width unchanged. This is an illustrative test region, not a recommended setting for any particular camera. Keep exposure, gain and the target position unchanged where the documented workflow permits.

Compare both dimensions and visible coverage. A result with unchanged dimensions may contain blank margins, duplicated rows or a scaled crop rather than additional scene information.

Interpret the observations as follows:

• Raw texture remains full-size; rectified texture shrinks: a retained source image may support another processing pass.
• Raw and rectified texture both shrink: investigate sensor readout, transport cropping and coupled ROI settings.
• Texture remains full-size; depth shrinks: the existing texture output may already meet the inspection requirement.
• Only rectified texture is accessible: internal sensor readout size does not establish that your application can retrieve more pixels.

Change only one documented setting at a time. If the software offers a choice between sensor cropping and processing-only cropping, compare those modes separately.

Verify that the retained image is the texture exposure

3. Confirm that the full-size buffer belongs to the intended texture exposure, not merely to another image in the scan sequence.

Raw does not mean pattern-free. A raw depth input can still contain projected illumination, while a rectified texture image can come from a separate exposure without that pattern.

Inspect the image content and available acquisition metadata. Confirm the relevant illumination state through the documented sequence controls. Do not assume that a stream name, matching dimensions or proximity in memory establishes exposure identity.

Use a stationary target first. Then repeat with representative motion to determine whether the texture image and depth result describe sufficiently similar part positions. A separate texture exposure can introduce displacement even when both outputs belong to the same scan operation.

Also evaluate contrast, saturation and focus for the actual inspection. An exposure suitable for coloring a point cloud may not resolve a small printed mark or surface defect.

If the inspection combines texture coordinates with depth, the raw versus rectified registration guide explains why matching image dimensions alone does not establish correspondence.

Reprocess retained images without another exposure

4. Use a second processing pass only when the software documents processing from retained raw inputs independently of acquisition.

A supported sequence can separate the required outputs:

• Acquire and retain the full-size texture source and required depth inputs.
• Process the restricted depth region.
• Rectify the retained texture source into the required larger output region, or inspect the raw texture directly.

This avoids another exposure only if the processing operation consumes existing buffers. A command that silently reacquires images does not meet that requirement. Check capture counters or documented timestamps before and after processing.

Raw-image inspection may be sufficient for presence checks or classification that do not require rectified coordinates. For dimensional measurements, geometric overlays or depth correspondence, account for lens distortion and coordinate mapping.

Full-size raw input also does not guarantee valid rectified pixels everywhere. Rectification can require source pixels outside the sensor boundary, and its valid output region depends on calibration and the requested mapping.

Follow the stored-calibration rectification workflow for the processing checks. If output dimensions or origins change, consult the intrinsic parameter conversion guide rather than reusing a cropped image's calibration unchanged.

Calculate the cost of retaining full-frame data

5. Estimate memory and transport requirements before making full-size retention the production default.

The following example describes hypothetical unpacked monochrome images, not a catalog product. Assume a full image of 1920 × 1200 pixels and a cropped image of 1920 × 400 pixels, both using one byte per pixel.

Retaining two full images requires 4,608,000 bytes per acquisition before padding, metadata or additional processing buffers. At 10 acquisitions per second, that represents 46.08 MB/s of image payload. It is not a prediction of processing throughput or interface performance.

A packed color image using three bytes per pixel would triple the corresponding payload. Rectified copies, temporary remapping buffers and queued acquisitions add further memory demand.

Processing-only cropping may reduce matching work without reducing camera transfer traffic. Sensor cropping can reduce transferred data but removes the opportunity to inspect omitted regions later.

Keep the acquisition buffers alive until both depth and texture consumers finish. For asynchronous pipelines, the camera frame memory lifetime guide covers the ownership problem behind stale or overwritten images.

Control parameter state and verify repeatability

6. Treat acquisition state and processing state as separate, recorded configurations wherever the interface supports that distinction.

Avoid changing a shared ROI while another thread is acquiring or rectifying images. Otherwise, a frame can be captured with one configuration and processed under assumptions from another.

A practical state record includes the effective sensor region, processing region, texture mode, exposure, gain, pixel format, calibration identity and frame association. Read back settings after applying them when the interface supports it. Requested values can differ from accepted values because of alignment or operating-mode constraints.

If sequential parameter changes are necessary, serialize the documented operations. Preserve the original state and restore it after the alternate processing pass, including error paths. Do not assume that a successful processing call also restored acquisition settings.

Verify the workflow across a restart and reconnection. Use a declared validation run, such as 100 consecutive acquisitions, checking for missing buffers, unexpected dimensions and incorrect frame associations. This is a functional check, not proof of a rare-failure rate.

Choose a documented fallback when retention is unavailable

7. Select the fallback according to which requirement cannot be relaxed: full coverage, timing, bandwidth or cycle time.

If the interface exposes only cropped texture images, first check whether it supports full-size acquisition with a smaller depth-processing region. Otherwise, use a documented separate texture capture or a dedicated inspection camera. Neither should be described as a no-additional-capture solution.

For a projected structured-light candidate, review the MV-DPS200P-02. Confirm raw texture access, processing independence and frame association in its current documentation before selection; those capabilities are not established by the product category.

If independent inspection imaging is required, consider the LS-CMR-0166 as a separate acquisition candidate. Validate its optics, exposure timing and motion behavior against the task rather than assuming compatibility from resolution alone.

Compare 3D sensors and area scan cameras through the camera finder. A dedicated camera adds calibration and timing work, but it can separate conflicting inspection and depth requirements.

Acceptance checklist before committing the workflow

Accept the retained-image approach only after checking all of the following:

• The source contains the required scene coverage, not an enlarged crop.
• The retained buffer comes from the intended texture exposure.
• Processing does not trigger an undocumented replacement capture.
• Calibration matches the image dimensions, origin and coordinate convention.
• Frame association remains correct under representative motion.
• Buffer ownership remains valid until every consumer finishes.
• Restart and recovery preserve the intended settings.
• Memory, transfer load and cycle time meet the application's limits.

Before ordering, establish image availability with documented outputs or representative sample captures. Start with the camera configuration workflow, or send your output requirements, image metadata and timing constraints through application support.