Whiteboard image filter in Mathematica

[luisvasquezperslearn (Luis Antonio Vasquez Reina)]

I came up with the following Whiteboard filter inspired by this github post. A Whiteboard filter is useful for making handwritten notes more readable.

whiteboardfilter[img_] := MinFilter[
  ImageApply[
   (If[Mean[#] > 0.85, 1, Min[#]]) &
   ,
   Sharpen[ColorNegate[
     GaussianFilter[MedianFilter[img, 2], 40] - img
     ],
    5
    ]
   ],
  1
  ]

More compactly:

whiteboardfilter[img_] := MinFilter[ImageApply[
   If[Mean[#] > 0.8, 1, Min[#]] &,
   Sharpen[ColorNegate[GaussianFilter[MedianFilter[img, 2], 40] - img], 5]
   ], 1]

Sample uses

This is a scan of some of my handwritten notes in black ink, blue ink, pencil, and red ink.

Applying the filter:

This is a photo of the same handwritten notes, taken with my tablet:

Applying the filter:

This is a photo of an actual whiteboard where I wrote with a red marker, a blue marker, a black marker, and a green marker, taken with my tablet.

Applying the filter:

Details

• Tweaking the parameters may help you get a better result for you. In my case, moving the bound 0.85 between 0.8 and 0.95 gave other acceptable results.
• I have not tested this filter with pictures of chalk over a greenboard or a blackboard.
• Though it can be easily inferred from the code, this is how the filter works:

1. Apply a Median filter with radius 2

2. To this, apply a Gaussian filter with radius 40

3. To this, subtract the original image, and color negate the result.

4. Sharpen this with radius 5.

5. Apply the following function to the channel values of each pixel: If the mean of the channels is greater than 0.85, put 1 (White), otherwise use the mean. This step helps a lot with the noise in the background of the images, because it replaces very clear pixels with the color White.

6. Apply a Min filter with radius 1.

---

[rohitnamjoshi]

Hi Luis,

Nice work. Another approach is described in this post.