bitmapfilter – Convolve an image with a kernel
Available on these boards
- bitmapfilter.morph(bitmap: displayio.Bitmap, weights: Sequence[int], mul: float | None = None, add: float = 0, mask: displayio.Bitmap | None = None, threshold: bool = False, offset: int = 0, invert: bool = False) displayio.Bitmap
The name of the function comes from OpenMV. ImageMagick calls this “-morphology” (“-morph” is an unrelated image blending algorithm). PIL calls this “kernel”.
For background on how this kind of image processing, including some useful
weightsvalues, see wikipedia’s article on the subject.The
bitmap, which must be in RGB565_SWAPPED format, is modified according to theweights. Then a scaling factormuland an offset factoraddare applied.The
weightsmust be a sequence of integers. The length of the tuple must be the square of an odd number, usually 9 and sometimes 25. Specific weights create different effects. For instance, these weights represent a 3x3 gaussian blur:[1, 2, 1, 2, 4, 2, 1, 2, 1]mulis number to multiply the convolution pixel results by. IfNone(the default) is passed, the value of1/sum(weights)is used (or1ifsum(weights)is0). For most weights, his default value will preserve the overall image brightness.addis a value to add to each convolution pixel result.mulbasically allows you to do a global contrast adjustment and add allows you to do a global brightness adjustment. Pixels that go outside of the image mins and maxes for color channels will be clipped.If you’d like to adaptive threshold the image on the output of the filter you can pass
threshold=Truewhich will enable adaptive thresholding of the image which sets pixels to one or zero based on a pixel’s brightness in relation to the brightness of the kernel of pixels around them. A negativeoffsetvalue sets more pixels to 1 as you make it more negative while a positive value only sets the sharpest contrast changes to 1. Setinvertto invert the binary image resulting output.maskis another image to use as a pixel level mask for the operation. The mask should be an image the same size as the image being operated on. Only pixels set to a non-zero value in the mask are modified.kernel_gauss_3 = [ 1, 2, 1, 2, 4, 2, 1, 2, 1] def blur(bitmap): """Blur the bitmap with a 3x3 gaussian kernel""" bitmapfilter.morph(bitmap, kernel_gauss_3, 1/sum(kernel_gauss_3))
- class bitmapfilter.ChannelScale(r: float, g: float, b: float)
A weight object to use with mix() that scales each channel independently
This is useful for global contrast and brightness adjustment on a per-component basis. For instance, to cut red contrast in half (while keeping the minimum value as black or 0.0),
reduce_red_contrast = bitmapfilter.ChannelScale(0.5, 1, 1)
Construct a ChannelScale object
The
rparameter gives the scale factor for the red channel of pixels, and so forth.
- class bitmapfilter.ChannelScaleOffset(r: float, r_add: float, g: float, g_add: float, b: float, b_add: float)
A weight object to use with mix() that scales and offsets each channel independently
The
r,g, andbparameters give a scale factor for each color component, while ther_add`, ``g_addandb_addgive offset values added to each component.This is useful for global contrast and brightness adjustment on a per-component basis. For instance, to cut red contrast in half while adjusting the brightness so that the middle value is still 0.5:
reduce_red_contrast = bitmapfilter.ChannelScaleOffset( 0.5, 0.25, 1, 0, 1, 0)
Construct a ChannelScaleOffset object
- class bitmapfilter.ChannelMixer(rr: float, rg: float, rb: float, gr: float, gg: float, gb: float, br: float, bg: float, bb: float)
A weight object to use with mix() that mixes different channels together
The parameters with names like
rbgive the fraction of each channel to mix into every other channel. For instance,rbgives the fraction of blue to mix into red, andgggives the fraction of green to mix into green.Conversion to sepia is an example where a ChannelMixer is appropriate, because the sepia conversion is defined as mixing a certain fraction of R, G, and B input values into each output value:
sepia_weights = bitmapfilter.ChannelMixer( .393, .769, .189, .349, .686, .168, .272, .534, .131) def sepia(bitmap): """Convert the bitmap to sepia""" bitmapfilter.mix(bitmap, sepia_weights) mix_into_red = ChannelMixer( 0.5, 0.25, 0.25, 0, 1, 0, 0, 1, 0)
Construct a ChannelMixer object
- class bitmapfilter.ChannelMixerOffset(rr: float, rg: float, rb: float, r_add: float, gr: float, gg: float, gb: float, g_add: float, br: float, bg: float, bb: float, b_add: float)
A weight object to use with mix() that mixes different channels together, plus an offset value
The parameters with names like
rbgive the fraction of each channel to mix into every other channel. For instance,rbgives the fraction of blue to mix into red, andgggives the fraction of green to mix into green. Ther_add,g_addandb_addparameters give offsets applied to each component.For instance, to perform sepia conversion but also increase the overall brightness by 10%:
sepia_weights_brighten = bitmapfilter.ChannelMixerOffset( .393, .769, .189, .1 .349, .686, .168, .1 .272, .534, .131, .1)
Construct a ChannelMixerOffset object
- bitmapfilter.mix(bitmap: displayio.Bitmap, weights: ChannelScale | ChannelScaleOffset | ChannelMixer | ChannelMixerOffset, mask: displayio.Bitmap | None = None) displayio.Bitmap
Perform a channel mixing operation on the bitmap
This is similar to the “channel mixer” tool in popular photo editing software. Imagemagick calls this “-color-matrix”. In PIL, this is accomplished with the
convertmethod’smatrixargument.The
bitmap, which must be in RGB565_SWAPPED format, is modified according to theweights.The
weightsmust be one of the above types:ChannelScale,ChannelScaleOffset,ChannelMixer, orChannelMixerOffset. For the effect of each different kind of weights object, see the type documentation.After computation, any out of range values are clamped to the greatest or smallest valid value.
maskis another image to use as a pixel level mask for the operation. The mask should be an image the same size as the image being operated on. Only pixels set to a non-zero value in the mask are modified.
- bitmapfilter.solarize(bitmap: displayio.Bitmap, threshold: float = 0.5, mask: displayio.Bitmap | None = None) displayio.Bitmap
Create a “solarization” effect on an image
This filter inverts pixels with brightness values above
threshold, while leaving lower brightness pixels alone.This effect is similar to an effect observed in real life film which can also be produced during the printmaking process
PIL and ImageMagic both call this “solarize”.
- bitmapfilter.LookupFunction
Any function which takes a number and returns a number. The input and output values should be in the range from 0 to 1 inclusive.
- bitmapfilter.ThreeLookupFunctions
Any sequenceof three
LookupFunctionobjects
- bitmapfilter.lookup(bitmap: displayio.Bitmap, lookup: LookupFunction | ThreeLookupFunctions, mask: displayio.Bitmap | None) displayio.Bitmap
Modify the channels of a bitmap according to a look-up table
This can be used to implement non-linear transformations of color values, such as gamma curves.
This is similar to, but more limiting than, PIL’s “LUT3D” facility. It is not directly available in OpenMV or ImageMagic.
The
bitmap, which must be in RGB565_SWAPPED format, is modified according to the values of thelookupfunction or functions.If one
lookupfunction is supplied, the same function is used for all 3 image channels. Otherwise, it must be a tuple of 3 functions. The first function is used for R, the second function for G, and the third for B.Each lookup function is called for each possible channel value from 0 to 1 inclusive (64 times for green, 32 times for red or blue), and the return value (also from 0 to 1) is used whenever that color value is returned.
maskis another image to use as a pixel level mask for the operation. The mask should be an image the same size as the image being operated on. Only pixels set to a non-zero value in the mask are modified.
- bitmapfilter.false_color(bitmap: displayio.Bitmap, palette: displayio.Palette, mask: displayio.Bitmap | None) displayio.Bitmap
Convert the image to false color using the given palette
In OpenMV this is accomplished via the
ironbowfunction, which uses a default palette known as “ironbow”. Imagemagic produces a similar effect with-clut. PIL can accomplish this by converting an image to “L” format, then applying a palette to convert it into “P” mode.The
bitmap, which must be in RGB565_SWAPPED format, is converted into false color.The
palette, which must be of length 256, is used as a look-up table.Each pixel is converted to a luminance (brightness/greyscale) value in the range 0..255, then the corresponding palette entry is looked up and stored in the bitmap.
maskis another image to use as a pixel level mask for the operation. The mask should be an image the same size as the image being operated on. Only pixels set to a non-zero value in the mask are modified.
- bitmapfilter.BlendFunction
A function used to blend two images
- bitmapfilter.BlendTable
A precomputed blend table
There is not actually a BlendTable type. The real type is actually any buffer 4096 bytes in length.
- bitmapfilter.blend_precompute(lookup: BlendFunction, table: BlendTable | None = None) BlendTable
Precompute a BlendTable from a BlendFunction
If the optional
tableargument is provided, an existingBlendTableis updated with the new function values.The function’s two arguments will range from 0 to 1. The returned value should also range from 0 to 1.
A function to do a 33% blend of each source image could look like this:
def blend_one_third(a, b): return a * .33 + b * .67
- bitmapfilter.blend(dest: displayio.Bitmap, src1: displayio.Bitmap, src2: displayio.Bitmap, lookup: BlendFunction | BlendTable, mask: displayio.Bitmap | None = None) displayio.Bitmap
Blend the ‘src1’ and ‘src2’ images according to lookup function or table ‘lookup’
If
lookupis a function, it is converted to aBlendTableby internally calling blend_precompute. If a blend function is used repeatedly it can be more efficient to compute it once withblend_precompute.If the mask is supplied, pixels from
src1are taken unchanged in masked areas.The source and destination bitmaps may be the same bitmap.
The destination bitmap is returned.