Previewing, Comparing and Publishing the Processed Image

Image Processing lets you examine a scientific image after intensity, color and contrast filters have been applied. In the GX-TXT Curve Digitalizer you can compare Original and Processed, navigate the two views, keep a filter pipeline in the configuration and, if the result is satisfactory, publish a new canonical Image Dataset.

It is essential to distinguish four objects: the source image, the filter sequence, the temporary preview and the Dataset that may be published. Curve digitizing always uses the coordinates of the original raster, while the filters change only its appearance.

1. Starting from a Verifiable Original Source

The example figure contains four colored synthetic traces, a grid and slight noise. It measures 1280 × 900 pixels and was loaded into the Viewer as the original Image Dataset.

Viewer showing the original synthetic graph with four colored traces.
Figure 1. The original scientific raster is displayed before any new ImageMagick preview is generated.
Close-up of the colored traces and grid in the original raster.
Figure 2. The original color reference remains the point of comparison for every photometric processing.

The four-corner calibration maps X to 0–10 s and Y to −20–80 mV. A series also contains a manual point: it will be used to check that pixel operations are not mistaken for changes to the numerical points.

Image Processing panel with display options and the Apply processing / Preview button.
Figure 3. The display options are not the same as Apply or publishing a Dataset.

2. Where Apply and the Other Commands Are

Open Image Processing from the tab bar or from View → Image Processing…. The filter rows are configured on the left side. The Apply processing / Preview command computes the result, while Show processed preview chooses which version to show.

Compare original and processed images (split view), Synchronize pan and zoom and Show digitized curve overlays in comparison are used for visual assessment. Publish image.raster Dataset… is instead a separate choice that creates a new canonical object.

3. Apply Without Active Filters

If no operation is enabled, pressing Apply processing / Preview does not produce a new image. The Viewer reports that there are no operations to apply.

Message stating that no processing operations are selected.
Figure 4. With no operation, the Viewer does not produce a new preview and reports why.
Filter list with all filters disabled.
Figure 5. At least one real filter must be enabled before a result can be computed.

This is the expected behavior: you first need to choose at least one filter with its checkbox.

4. First Preview: Grayscale

Enable Grayscale and, for this example, turn off Live preview after changing a control so that the calculation must be requested explicitly.

Grayscale enabled with Live preview off and the original image still shown.
Figure 6. The pipeline has been changed, but the Canvas shows the previous image until Apply is run.

Press Apply processing / Preview. When processing ends, the GUI confirms Ready: 1280 × 900 px; raster geometry unchanged.

Processed preview in grayscale with the Ready status message.
Figure 7. The color image was converted to gray levels while keeping 1280 by 900 pixels.
Close-up of the grayscale traces in the processed preview.
Figure 8. Photometric changes can make a curve easier or harder to distinguish.

The curves change their tonal rendering, but not their position or size. In particular, conversion to gray can bring two traces closer together that were originally easy to tell apart by color.

5. Original and Processed: Switching Without Deleting

Disabling Show processed preview brings back the original color image.

Original color image shown with Show processed preview turned off.
Figure 9. The processed preview still exists, but the screen shows the source colors again.

Re-enabling the checkbox shows the processed version again.

Processed grayscale preview shown again.
Figure 10. Switching changes the display without replacing the source Dataset.

The button does not recreate the original Dataset, does not apply a geometric transformation and does not accept any scientific point. It is a display choice between two images available in the Viewer.

6. Turning On the Side-by-Side View

Select Compare original and processed images (split view). The Canvas is split into two panels: Original on the left and Processed on the right.

Split view with Original on the left and Processed on the right.
Figure 11. The left view shows the color source and the right one the gray preview.
Close-up of the split view showing matching features.
Figure 12. The same raster features can be recognized in both views, with no change to the geometry.

This comparison is particularly useful with several curves, dense grids or faint features: a line made more visible on the right must still be recognizable and consistent in the original on the left.

7. Showing and Hiding the Digitized Points

With Show digitized curve overlays in comparison, the scientific markers are available in both panels. Disabling the function makes the markers disappear from the view.

Split view with curve overlays turned off.
Figure 13. The Show digitized curve overlays command hides the markers without deleting the data.
Close-up of the split view without point markers.
Figure 14. The visual traces remain; only the overlay of the annotations changes.

Re-enabling the command, they become visible again.

Split view with the manual point marker shown in both panels.
Figure 15. The manual point returns in both views as an indicator, not as an operation on the data.

The manual point in the example was not deleted: the function controls the graphic overlay, not the records of the series.

8. Independent Zoom and Pan

With Synchronize pan and zoom disabled, each panel has its own navigation. Use the mouse wheel to zoom into the Original side.

Split view with only the Original panel zoomed in.
Figure 16. The Processed view keeps its previous zoom while Original increases its scale.

The Processed side keeps its previous level. Likewise, dragging with the right or middle button moves the Original view without dragging the other one.

Split view with only the Original panel panned.
Figure 17. Moving the source raster does not drag the Processed panel when Sync is Off.

This choice is useful when you want to examine a detail at a different magnification, but makes point-by-point comparison less immediate.

9. Synchronizing the Two Views

Enable Synchronize pan and zoom. The Viewer aligns the two panels on the same position and scale of the source.

Split view aligned after enabling Synchronize pan and zoom.
Figure 18. The Viewer aligns the two images on the same raster scale and position.

Zooming into the Processed panel now updates Original as well.

Both panels zoomed in together.
Figure 19. The two views update their zoom level in a coordinated way.
Close-up of the same region in both synchronized panels.
Figure 20. The operator can directly compare the same region in the original and in the filtered image.

With synchronization active, panning one panel is also reflected in the other.

Both panels panned together with synchronization on.
Figure 21. Moving the Processed view updates the position of Original when Sync is active.

The alignment follows the coordinates of the source raster, not the apparent identity of the lines after processing.

10. Moving the Central Divider

The vertical divider between Original and Processed can be dragged to give more space to one panel.

Split view with the divider moved to one side.
Figure 22. The two areas can have different widths while keeping the reference to the source coordinates.
Split view with a wider Processed panel.
Figure 23. An asymmetric split gives more space to the image that needs a closer check.

The two areas can therefore have different widths, while synchronization keeps using the same region of the raster. Dividing the space changes neither the size nor the calibration of the images.

11. Back to the Single Canvas and the Curves Tab

Disable Compare original and processed images to return to the single Canvas.

Single Canvas view after turning off the comparison.
Figure 24. The Viewer returns to the usual single view without deleting the processed pipeline.

Then open Curves.

Curves tab with the processed preview visible on the Canvas.
Figure 25. The processed view can help to read the curve, but the points remain referred to the original raster.

The processed version can stay visible as an aid to digitizing, but the points added from the Curves tab refer to the original coordinates. Going back to Image Processing, the Viewer keeps the single-view choice.

Image Processing tab with the single-image layout.
Figure 26. Turning the comparison off keeps the single-image layout.

The Image Processing tab must not be used as a point editor: the comparison view neither adds nor deletes scientific points.

12. Live Preview: Automatic Update

Turn Live preview after changing a control back on. Add Gamma to the previous Grayscale operation, setting for example Gamma = 1.45.

Status message reporting that the pipeline has changed.
Figure 27. The Viewer reports that the pipeline has been changed and must update the result.
Processing status during the automatic recalculation.
Figure 28. With Live preview enabled, recalculation starts without pressing Apply manually.

The GUI first reports that the pipeline has changed; then it automatically starts the calculation and shows a new preview consistent with the updated parameters, without having to press Apply again.

Ready status message after the live preview update.
Figure 29. The Viewer reports Ready after generating the version consistent with the new parameters.

The update is briefly deferred when sliders or fields are changed: this limits unnecessary recalculations during adjustment. Having a preview ready does not mean the result has already been published.

13. Try Automatic Enhancement

The Try automatic enhancement button analyzes the source and proposes a sequence of filters. In the demonstration figure Auto levels and Local contrast (CLAHE) were suggested.

Filter list with Auto levels and Local contrast (CLAHE) enabled by the suggestion.
Figure 30. The automatic suggestion selects Auto levels and CLAHE, to be evaluated before any publishing.
Detail of the suggested filter settings.
Figure 31. The automatic proposal is a sequence to check, not a certificate of scientific correctness.

The Viewer can compute the preview of the automatic proposal.

Processed preview produced by the automatic suggestion.
Figure 32. The operator must compare the proposal with the source and check that no useful stretches were lost.

It is a practical suggestion, not a scientific check. It can improve the readability of a line or introduce unwanted effects on faint traces, intersections or edges.

14. Restoring the Previous Pipeline

Restore filters before auto enhancement recovers the sequence used before the automatic suggestion, as long as it remains valid for the current source and configuration.

Filter list with Grayscale and Gamma restored.
Figure 33. Restore filters before auto enhancement brings back Grayscale and Gamma with their previous values.
Detail of the restored Gamma value.
Figure 34. Restoring keeps the distinction between automatic suggestion and choice of method.

In our case Grayscale and Gamma 1.45 come back. The next preview uses those filters again.

Processed preview with Grayscale and Gamma.
Figure 35. The new image is again consistent with Grayscale and Gamma instead of the suggested filters.

It is advisable to compare both alternatives on the original image before adopting an automatic processing.

15. When a Preview Becomes Out of Date

Suppose Gamma is changed or Median is added after a preview was already ready, with Live preview kept disabled.

Filter list changed after the last preview.
Figure 36. A change to Gamma or Median makes the previously displayed result out of date.

The old visible result does not necessarily represent the new pipeline. If you try to publish it right away, the Viewer blocks the operation and requires a new Apply processing / Preview.

Message asking to apply the current filters before publishing.
Figure 37. The GUI requires the current filters to be applied before allowing a new Dataset.
Detail of the out-of-date preview message.
Figure 38. The safeguard prevents publishing pixels derived from parameters different from the current ones.

This check avoids associating with a Dataset a list of filters different from the one that actually produced the pixels.

16. Preparing the Pipeline to Publish

For the final example we choose three operations in the Standard sequence:

PositionOperationParameter
1GrayscaleOn
2Median denoise2
3Gamma1.6

Press Apply processing / Preview and check the Ready message.

Processed preview with Grayscale, Median 2 and Gamma 1.6 and the Ready status.
Figure 39. The new 1280 by 900 pixel preview is ready with three operations in the intended order.

This choice changes colors and intensities but does not resize, rotate, flip or distort the raster.

17. The Output ID Cannot Be the Same as the Source ID

In the Publish processed Image Dataset section, the Image Dataset ID field identifies the new object.

If the same ID as the source is entered and Publish is pressed, an error appears.

Error message for an output ID equal to the source ID.
Figure 40. A new Image Dataset cannot overwrite the identity of the original raster.
Detail of the Image Dataset ID field with the source ID.
Figure 41. The check protects the canonical identity of the source in the Session.

The Viewer thus protects the identity of the original Dataset; the processed image must be published separately.

18. Choosing the ID and Description of the Result

For the example two distinct IDs are used, visible in the Dataset selector:

RoleImage Dataset ID
Starting imagev20.source.image
Filtered imagev20.processed.image

The Image description can state the content, for example Processed V20 analytical color graph. The names are only example identifiers: in your own Session choose clear IDs consistent with the scientific project.

Publish processed Image Dataset section with ID and description filled in.
Figure 42. Before publishing, check the name, the ID and the sequence of operations.

The description accompanies the derived object; the identifier remains its stable canonical identity.

19. Confirming No: Keeping Only the Preview

Press Publish image.raster Dataset…. The Viewer shows a confirmation request with the output Dataset and the source used.

Publish confirmation dialog with output and source Dataset.
Figure 43. The message summarizes the derived Dataset and the source; the operator can cancel.
Detail of the Yes and No buttons in the confirmation dialog.
Figure 44. The canonical operation is performed only after explicit confirmation.

Choosing No, the new Dataset is not created.

Image Processing panel after choosing No.
Figure 45. The result stays available as a preview, without creating an additional canonical Dataset.

The preview stays available in the Viewer and can still be reviewed or changed. Cancelling the publication does not necessarily cancel the filter settings.

20. Confirming Yes: Publishing a New Image Dataset

Repeating Publish and choosing Yes, the Viewer sends the processed raster to the GX-TXT canonical publishing system.

Publish confirmation dialog before choosing Yes.
Figure 46. Choosing Yes afterwards authorizes the registration of a new, independent Image Dataset.

When the operation ends, the successful publication status appears.

Successful publication status for v20.processed.image.
Figure 47. The v20.processed.image Dataset is registered with its own SHA, distinct from the original.
Detail of the publication message with the SHA-256 prefix.
Figure 48. The message shows the new canonical identifier and the beginning of its SHA-256.

The same Session now contains two Image Datasets: the original and the derived one. Publishing does not change the ID of the first or turn the scientific points of the curve into a separate numerical Dataset.

21. What Is Actually Published

The object created is a standalone image.raster containing the processed raster and a set of provenance metadata. It is not simply the Viewer’s internal preview.

FeatureSourceProcessed Dataset
IdentityOriginalNew ID
Dimensions1280 × 9001280 × 900
PixelOriginal colorsProcessed gray levels
SHA-256Original identifierDifferent SHA-256
Geometric calibrationOriginal referenceUndistorted pixel mapping
Numerical tracesBelong to the configurationNot published with the image

A processed Dataset can therefore be opened independently, but must not be confused with publishing curves or their uncertainties.

22. Provenance Information Kept

The derived Dataset keeps a manifest including the identifier and SHA-256 of the source, the ImageMagick version, the dimensions, the filter order, the parameters, the unchanged-geometry flag and a fingerprint of the processing.

In the example, the documented order is Grayscale → Median 2 → Gamma 1.6. Changing that order can produce different pixels, even if the individual operations stay the same.

To reconstruct a processing over time you therefore need both the source raster and the exact manifest of the filters, not just the caption of an image.

23. Numerical Check of the Processed Raster

In the example, the pixels of the published Dataset were compared with a second ImageMagick processing run on the same characteristics of the original.

All 1,152,000 pixels underwent at least one change in the RGB channels compared with the source; the result is in grayscale. Repeating the processing with the same parameters produced zero pixel differences from the new Dataset.

This demonstrates the technical reproducibility of this example. It does not mean that the filters increase scientific accuracy, and it is not an estimate of metrological uncertainty.

24. Save, Close and Reopen After Publishing

Save the Viewer configuration with Configuration → Save configuration, close the Viewer normally and reopen it from the same Session.

Reopened Image Processing tab with the saved pipeline.
Figure 49. The Viewer recovers source and pipeline and can regenerate the same processed preview.

The Grayscale, Median 2 and Gamma 1.6 operations stay recorded; the GUI can regenerate the same preview when ImageMagick is available. The two Datasets keep appearing in the Session catalog.

25. Opening the Processed Dataset in a New Configuration

A published Dataset is not tied exclusively to the preview it came from. You can create a new Curve Digitalizer configuration, choosing the processed version as the Image Dataset.

New configuration created with the processed Image Dataset.
Figure 50. The published raster can become an independent source without replacing the original document.
New configuration showing the grayscale raster and no points.
Figure 51. The new document starts with no numerical points, leaving the first configuration unchanged.

In the new configuration the processed image becomes the source; the scientific points acquired in the original configuration are not copied automatically.

26. Going Back to the Original Configuration

Use the Configuration selector to go back to the document that created the processing.

Original configuration with the color raster and the manual point.
Figure 52. The initial document keeps the scientific point and its reference to the color raster.
Image Dataset selector listing both the source and the processed Dataset.
Figure 53. Original source and processed Dataset remain available in the same Session.

The source becomes the original raster again, with its own calibration and with the manual point already stored. The existence of the derived image neither deletes nor overwrites the initial project.

27. Persistence of the Two Configurations

After another close and reopen, you can still choose the configuration that uses the processed image.

Reopened configuration using the processed image.
Figure 54. The document linked to the new raster survives closing and reopening the Viewer.

You can then go back to the original one.

Reopened original configuration with the manual point and calibration.
Figure 55. The manual point, the calibration and the original pipeline remain available together with the new Dataset.

The two configurations therefore represent two different workflows: one keeps the direct reference to the source and to the points already acquired; the other lets you start an independent tracing on the filtered Dataset.

28. The Scientific Points Are Not Changed by the Comparison

Comparison, zoom, pan, overlays and preview selection must not be interpreted as commands to add, move or delete points.

In our example, the manual point of the original configuration keeps its identity, pixel coordinates and scientific coordinates throughout processing, publishing and reopening.

Points are edited in the Curves tab with the appropriate commands, not through the Image Processing Canvas.

29. Publishing the Image Does Not Publish the Curves

Publish image.raster Dataset… registers an image with its own metadata. It does not automatically create a Dataset of curve coordinates.

To make the numerical results available, follow the controls of the Curve Digitalizer Publish tab and explicitly choose the series or scientific elements to publish.

The two processes have distinct meanings: one produces a derived raster, the other produces numerical scientific data.

30. Invalid Preview, Delay and Current Source

When the filter, value, order or source changes, the previous preview can become out of date. Before publishing, the Viewer requires the new processing to match the current source and parameters exactly.

With Live preview enabled the recalculation happens automatically, after a short interval; with Live preview disabled you need to press Apply.

The Ready text must be associated with the last pipeline actually computed, not with an old image still present on the Canvas.

31. Appropriate Use of the Automatic Suggestion

Try automatic enhancement can be useful as a starting point for dark scans, uneven contrast or noise. However, do not assume that the proposal is better for every curve.

Always compare the result with Original, especially if a thin line, an intersection or a faint trace risks disappearing.

Restore filters before auto enhancement helps recover the method chosen earlier, without losing control of the experiment.

32. If ImageMagick Is Not Available

The Viewer can still open the image and use the digitizing functions, but it cannot compute a new processed preview without the configured engine.

Check GX-TXT Configuration → Runtime → ImageMagick and the message shown in the Image Processing tab. Do not change the pipeline at random to solve a program availability error.

If processing fails, check the error message before trying again.

33. Watch Out for Bit Depth and Source Characteristics

The pipeline processes a decoded RGB representation of the source image and publishes an RGB PNG image. It is not a system designed to preserve, without transformation, the original depth of every channel of a high-precision acquisition.

For metrological work based on the real pixel intensities, the original Dataset must be kept and used as the authoritative reference.

Contrast, gamma, threshold and noise filters can change the apparent position of edges without changing the coordinate grid.

34. Common Errors and Checks

SituationMeaningCorrect action
No processing operations selectedNo active filterEnable a valid filter
Out-of-date previewThe filters changed after the last calculationPress Apply and wait for Ready
ImageMagick unavailableEngine not configuredCheck Runtime
Output ID equal to the sourceAn attempt to reuse a canonical identityChoose a different ID
Publication cancelled with NoNo new Dataset registeredKeep checking the preview
Automatically suggested operationProposed choice, not approvedCompare Original and Processed

A particularly important check concerns numeric filters: in the version examined, disabling and re-enabling some filters can reset their parameters to the defaults. After a toggle, check the value and the order again before the next Apply.

35. Recommended Workflow

  1. Load an identified original image and check its calibration.
  2. Prepare the filters in Standard or Advanced, justifying order and parameters.
  3. Use Apply processing / Preview or Live preview and wait for completion.
  4. Compare Original and Processed, possibly with synchronized zoom and pan.
  5. Check thin lines, intersections, loss of detail and possible artifacts.
  6. Check the parameters again after any disabling/re-enabling.
  7. Enter a distinct Image Dataset ID and a meaningful description.
  8. Confirm Publish image.raster Dataset… only after checking the current version.
  9. Check that both Datasets exist and that the original source is kept.
  10. Save the configuration and, if needed, open the processed image in a new document.

36. Summary

Curve Digitalizer keeps image preparation separate from point collection and from the publication of numerical results. Image Processing can provide a more readable version of the graph, while Original keeps the reference to the pixels and the calibration.

Side-by-side comparison, synchronized navigation, filter restoration and publishing confirmations make it possible to keep control of the path from the source image to the derived Dataset.

The decisive choice remains scientific: an image that is easier to trace is not automatically more accurate. Always keep the original and visually check any processing before using the digitized curve in further analyses.