Digitise image graphs and curve plots by following the procedure:

1. Load your graph image by Browse, Drag & drop or Paste the image file onto the page. JPG, PNG or GIF formats are accepted.
   Keep the image size within the max. limits.
2. Mode:
   Graphreader V2 offers 2 modes:
   In Standard mode the curve will follow a normal graph with increasing x-values.
   In Freeform 2D mode the curve is free form and can go in any direction.
   Note: Changing Mode may delete some of the inserted fixpoints.
3. Start by defining the axis scaling by reshaping the blue axis box. Align the box to well-defined lines and features on the graph image to get the most accurate scaling. Precise alignment is essential to get an accurate result.
   
   
   Hint: Use Snap to lines to help aligning the axis box to lines and edges on image.
   Note: The axis box does not define the data range thus no need to encircle the targeted curve segment. The sampled data range is defined by first and last fixpoint.
   If needed, the axis box can be rotated slightly by dragging (right-click) the corners of the axis box.
   
4. Enter corresponding axis coordinates for the 4 corners of the axis box.
   
   
   The coordinate system for the graph is now set.
5. Place fixpoints on targeted curve. Double click to create new fixpoints and right click to delete a fixpoint.
   
   
   For standard usage (linear sampling), you may manually draw up the curve by use of a high number of fixpoints.
   When using machine estimation of the curve, just place a few fixpoints on the curve. Make sure to place fixpoints where the curve is well defined against the background.
   Hint: Use Snap to lines to help placing the fixpoints center on the curve.
   
6. Sampling intervals: Before generating the curve, enter the sampling step length along the x-axis. The sampling values will be relative to the axis datum. If unchecking ‘Relative to axis zero’ the sampling steps will start from the first fixpoint.
   
   
   Alternatively the sampling can be defined in terms of number of samples between the first and last fixpoint.
   Hint: Activate to show the sampling grid locations.
   
7. Default sampling method is linear interpolation between the fixpoints.
   
   
   Instead of linear interpolation it is also possible to use various smooth spline interpolation functions, including: quadratic, cubic, cubic_natural, pchip and akima, each giving a unique smooth curve form.
   
8. When using Machine estimation the curve is tracked by use of image analysis.
   
   
   In this case you will normally only need to place 4-7 fixpoints on the curve to get a complete curve fit. Placing the fixpoints where the curve is well-deifined to the background and clear of other curves will help the machine estimation to get a better result.
   Hint: The machine estimation will not always work. Try placing a few more fixpoints on the problematic parts of the graph. You may also try the different available algorithms:
   Standard estimation will work for most cases where the target curve stands out with a distinct color.
   Thin curve may give a better result for graphs with very thin curves
   Monochromatic may give a better result for black/white graphs.
   
9. Hit to sample the curve to sample the curve according to the inserted fixpoints.and chosen sampling method.
   
   The sampled curve is show on the graph image when clicking
10. Data output: Scroll down to inspect the generated sampled data.
    
    
    
    JSON and CSV data can be readily be exported. A simple plot of the sampled data is also shown.
    
11. Accuracy. A rough estimate of the accuracy (best case) of the produced y-data assuming a +/- 0.5 pixel accuracy for all fixpoints and axis box. The slope of the curve is furthermore taken into account.
    
    
    This method is not available for logarithmic axis settings.
12. Post processing.
    
    
    A few simple post processing options are available in case you need to apply smoothing or regression to the sampled curve.
    Denoise: Experimental noise removing filter to eliminate smaller fluctuations on the curve (Beta)
    Spline fit: Approximating the curve with a spline of a given order
    Linear regression: Polynomial regression of given order.
    Moving average: A simpel moving average filter of given length. Symmetric endpoint conditions.
    Gauss filter: Applying a Gauss shape moving filter window of given length. Symmetric endpoint conditions.
    

Colorbar Type:

Indicate type of legend colorbar, having Continous or Discrete colors

About graphreader.com

Graphreader is meant to offer a set of simple online tools for extracting values from image based graphs and plots, and to export those data into comma seperated values (CSV) or JSON.

Graphs and curves are effective tools for presenting results in reports and publications, however they are also effective means of manupulation since it can be a challenge to grasp and compare the exact values. With graphreader it is time to go deeper into those data and carry out real quantitative analysis.

Obviously, the resulting values will at best be an approximation to the original data set, and you need to do your own data validation, especially when using the machine estimation feature. There may easily be slight offsets and systematic errors, however normally you will get a good idea of the actual data.

Data privacy

Uploaded data will not be used or shared in any way, however since I am not in full control of how data is stored and backed up on the hosted cloud service, please dont upload any confidential material.

Contact and Feedback

This tool is developed as a project in web-programming, data science and python.
Feel free to leave your feedback via the Feedback option or on .
You may check out my Blog on audio field recordings at www.recordingsofnature.wordpress.com
or at LinkedIn : www.linkedin.com/in/kristian-larsen-74006320