Neat Image - Guide de l'Utilisateur / Profils de Bruit de Dispositif / Building profile for specific device mode / Using a regular image / Analyzing image noise / Manual profiling

Sous-Étape B. Réglage fin d'analyse

Pour une réduction de bruit plus précise, il est utile de mesurer la dépendance entre le niveau de bruit et la luminosité locale dans des secteurs d'image différents. Cette dépendance devrait être prise en considération si le bruit de l'image dépend sensiblement de la luminosité (par exemple, si le bruit est fort dans des secteurs foncés et faible dans des zones claires).

Le Réglage fin d'analyse (Fine-Tuning Analyzer) mesure cette dépendance. Les résultats de mesures sont montrés par l'égaliseur dans la fenêtre Fine-Tuning Analyzer. L'égaliseur a neuf curseurs correspondant à la gamme de luminosité de la plus foncée à la plus plus claire pour chaque sonde (R, V, B) du dispositif d'acquisition d'image.

Les valeurs des curseurs de l'égaliseur correspondent aux niveaux de bruit estimés dans différentes gammes de luminosité relative au profil approximatif de bruit. Les valeurs positives des curseurs reflètent des niveaux de bruit plus élevés et font que Neat Image considère plus d'éléments comme étant du bruit; des valeurs négatives reflètent des niveaux de bruit plus bas et font que les éléments considérés comme étant du bruit sont moins nombreux dans les gammes de luminosité correspondants.

Le Réglage fin d'analyse (Fine-Tuning Analyzer) peut être employé de manière automatique ou manuelle. Ci-dessous, la méthode automatique est décrite en premier. La méthode manuelle est décrite en détail pour fournir une meilleure compréhension du processus et du résultat. La méthode manuelle n'est disponible que dans le mode avancé Advanced Mode (voir le menu Tools | Advanced Mode).

Cas du réglage fin automatique

Le réglage fin automatique trouve et analyse plusieurs secteurs de l'image sans traits distinctifs plats automatiquement. Vous n'avez pas besoin de faire quoi que ce soit manuellement; le réglage fin auto peut être fait juste d'un clic :

Cliquer sur le bouton (Auto Fine-Tuning Analyzer) ou sélectionner le menu Profile | Auto Fine-Tune.

The whole image will be automatically analyzed by Neat Image and some of the equalizer sliders will receive the 'measured' status (see diagrams details). The values of other sliders will be then automatically interpolated by the Auto Complete function and will receive the 'manual' status.

You may want to inspect the equalizer values after applying auto fine-tuning. In most cases, there is no need to do manual slider adjustments. If you feel this is necessary, please follow the guidelines of the manual fine-tuning subsection below. Otherwise please proceed to Étape II. Documenting the noise profile.

 

Case of manual fine-tuning

Manual fine-tuning involves analyzing several flat featureless image areas. You have to manually find and analyze them one after another:

1. Find and select a uniform featureless image area

• Scroll, pan, zoom the image in the viewer in the Device Noise Profile tab to find a uniform image area.
  
  
• Use the mouse to select a uniform image area: press the left button, drag the mouse and then release the button.
  
  See examples of uniform image areas.
  
  The size of an image area can be from 30x30 to 300x300 pixels. The selection frame will change its thickness according to the selection size. Also, when you are selecting an image area, the selection status on the bottom of the Fine-Tuning Analyzer box is dynamically indicating which frequency components (see what is frequency) are contained in the selected area and would be analyzed: 'high', 'high+mid', 'high+mid+low', 'high+mid+low+very low freqs'.
  
  

  Size of an area, pixels	Which frequency components would be analyzed	Rating
  200x200 - 300x300	High, medium, low and very low	Best
  100x100 - 200x200	High, medium and low	Good
  60x60 - 100x100	High and medium	Ok
  30x30 - 60x60	High	Poor

  
  The selected area would be analyzed according to its frequency composition (of high, medium, low and very low frequency image components). When a frequency component is not analyzed, all the data related to this component are estimated (extrapolated). That is always not accurate; therefore, it is best to choose large areas so that all the frequency components could be analyzed.
  
  

  Warning

  The selection status displays "signal clipping!"when the image in some of the channels (R,G,B) is close to the dynamic range limit in the selected area. Fine-tuning a device noise profile using the selected area could be inaccurate. Please try to avoid this for best results.

  
  When you select an image area, its position in the brightness range is shown with red font color of the value(s) of the corresponding slider(s)' in the noise profile equalizer. Also, it is displayed by the color indicators at the bottom of the equalizer.
  
  

2. Analyze selected image area with Manual Fine-Tuning Analyzer

• Click (the Manual Fine-Tuning Analyzer button) or select the Profile | Fine-Tune Using Selected Area menu item.
  
  The analysis results are shown in noise profile equalizer. For uniform areas with noise only, the corresponding slider receives the 'measured' status - the green shading on the slider's value, like . If an area with signal clipping has been used to analyze noise characteristics then the corresponding slider receives the 'inaccurate' status - the red shading, like . When an area with unexpectedly strong level of noise is encountered, the orange shading is applied, like .
  
  

  Warning

  The orange shading is applied when the analyzed noise in the corresponding brightness range is unexpectedly strong. There are several possible reasons for that: Fine-tuning is being done using a bad (e.g., containing visible details) image area; Wrong device noise profile is used (the profile's device and device mode do not match those of the analyzed image OR the rough noise analysis has been done inaccurately); Noise in this image is unusual and contains strong variations. The orange shading is a warning sign. It does not necessarily signify wrong measurement. Please make you own judgment in this situation and if necessary rebuild the device noise profile or select a more uniform area for fine-tuning. The red shading is a sure sign of wrong measurement. You have to reset the corresponding slider or undo the last analysis (see below).

  
  
  

3. If necessary, reset status of a slider (optional)

• Click on the color shading of a slider to reset its status and value.
  
  If a slider has red (or any other color) shading, you can safely reset it and analyze another part of the input image to re-measure its value. There is no need to reset the entire equalizer because of one wrong value.
  
  

4. If necessary, undo the last analysis (optional)

• Click (the Undo button) or select the Profile | Undo Last Fine-Tuning Analysis menu item.
  
  Undoing the last fine-tuning analysis may be useful when a bad choice of image area has resulted in bad analysis results.
  
  

5. If necessary, reset the whole equalizer (optional)

• Click (the Reset fine-tuning results button) or select the Profile | Reset Fine-Tuning Results menu item.
  
  

6. Repeat 1-5 with other uniform image areas of different brightness

To make a device noise profile more accurate you have to fine-tune it using several uniform areas of the image (naturally, analyzing the same area many times makes little sense). Try to choose uniform areas to cover all brightness ranges in all channels of the equalizer (i.e., to get shadings on all sliders' values). Use color shadings as well as red markings (which are used to reflect the range of the current selection; like -40%) to guide the process of fine-tuning. Also use the color indicators on the bottom of the Fine-Tuning Analyzer box as guidance when doing that. If the majority of sliders' values have green shadings, you can stop the process.

7. Set remaining sliders at your option (optional)

If not set by the Fine-Tuning Analyzer, the sliders of the equalizer have default values. You can leave them with default values or can adjust these sliders to bring them into better agreement with the measured ones. Adjusting the sliders can also be done automatically or manually:


• Manually adjust the remaining sliders as you consider necessary.
  Manually adjusted sliders receive the 'manual' status (a yellow shading, like ).
  
  or
  
  
• Use Auto Complete to automatically adjust the unmeasured sliders by interpolation based on the measured data. Click (the Auto Complete button) or select the Profile | Auto Complete menu item to automatically complete the fine-tuning.
  
  Using Auto Complete is highly advisable as the last step of the manual fine-tuning process.
  
  
  The figure below summarizes the status marking of sliders:
  
  
  
  
  When using the noise profile equalizer, use the color indicator to simplify the fine-tuning process. Colored lines of the indicator show:
  
  
• which sliders of the equalizer correspond to the colors of the selected image area/pixel (use the Shift key for pixel-wise indication);
• which slider values are different from default values;
• which sliders have (possibly) incorrect values.

The figure below explains each state of the indicator elements:






How to check if a device noise profile has been fine-tuned properly

The equalizer sliders should be mostly shaded in green and, occasionally, yellow. The color indicators should be filled with solid colored lines at all positions.

Sous-Étape A :: Sous-Étape B