Differences between revisions 70 and 71

Images and bitmaps in wxPython are normally created by loading files from disk. Working with the images in memory is, however, a common task for certain types of application. This recipe set looks at the in-memory manipulation (and particularly the generation) of images for use in wxPython applications.

Table of Contents:

Contents

Thanks

Formats and Conversions

The Image classes I find myself working with most frequently are wxImage, wxBitmap, and PIL Image. Here are functions I use to convert between them, in all combinations. I like to have these broken out as six individual functions with related names because the usage and naming in the wxWindows API can be confusing. The API also changes relatively frequently. I believe that the older text in this section is outdated, but I haven't had time to go through it. -- RobbShecter

Toggle line numbers

   1 # Tested with wxPython 2.3.4.2 and PIL 1.1.3.
   2 from wxPython import wx
   3 import Image             # Only if you need and use the PIL library.
   4 
   5 def bitmapToPil(bitmap):
   6     return imageToPil(bitmapToImage(bitmap))
   7 
   8 def bitmapToImage(bitmap):
   9     return wx.wxImageFromBitmap(bitmap)
  10 
  11 
  12 def pilToBitmap(pil):
  13     return imageToBitmap(pilToImage(pil))
  14 
  15 def pilToImage(pil):
  16     image = wx.wxEmptyImage(pil.size[0], pil.size[1])
  17     image.SetData(pil.convert('RGB').tostring())
  18     return image
  19 
  20 #Or, if you want to copy alpha channels too (available from wxPython 2.5)
  21 def piltoimage(pil,alpha=True):
  22    if alpha:
  23        image = wx.EmptyImage( *pil.size )
  24        image.SetData( pil.convert( "RGB").tostring() )
  25        image.SetAlphaData(pil.convert("RGBA").tostring()[3::4])
  26    else:
  27        image = wx.EmptyImage(pil.size[0], pil.size[1])
  28        new_image = pil.convert('RGB')
  29        data = new_image.tostring()
  30        image.SetData(data)
  31    return image
  32 
  33 
  34 def imageToPil(image):
  35     pil = Image.new('RGB', (image.GetWidth(), image.GetHeight()))
  36     pil.fromstring(image.GetData())
  37     return pil
  38 
  39 def imageToBitmap(image):
  40     return image.ConvertToBitmap()

Conversions among wxImage, wxBitmap, wxCursor, wxIcon and DATA
- wxImage to wxBitmap -- myWxImage.ConvertToBitmap() or wxBitmapFromImage(myWxImage) wxImage to DATA -- myWxImage.GetData() returning a string in width * height * 3 format DATA to wxImage -- image = wxImage(); image.SetData( data ) where data is a Python string of length width * height * 3. DATA to wxBitmap -- Go through wxImage to get to wxBitmap. DATA to wxIcon -- Should be possible, but I don't see an overloaded-constructor name for it. wxIcon to wxBitmap -- bitmap = wxEmptyBitmap( icon.GetWidth(), icon.GetHeight()); bitmap.CopyFromIcon( icon )
There seem to be some missing functions, not sure if these are available at a lower level of abstraction for the C peoples, but anyway, here's what hopefully will show up some day:
- wxBitmap to DATA or wxImage -- Probably the most critical missing item, currently you have to use "save to file" then load the bitmap back from disk using a wxImage. DATA or wxImage to wxCursor -- Should be an overridden constructor available, not yet. wxIcon to DATA or wxImage -- To allow for editing of icons, this would be required. wxCursor to DATA or wxImage -- Again, to allow for editing cursors. Would also need methods to get the current hot spot. wxImage to wxIcon -- See DATA to wxIcon above, either one would be fine, but would be nice to have the complete set.

PIL (Python Imaging Library)

PIL can be used with wxPython if image processing services are required beyond loading the formats built into wxPython. DATA formatted strings can be created from PIL Image objects using the .convert and .tostring methods (as show in the example below). Note the use of the size attribute of the PIL Image to create the empty wxImage object.

        def GetBitmap( self ):
                import Image
                source = Image.open( r"Z:\mcfport\portfoli\full\claire.jpg", 'r')
                image = wx.EmptyImage( *source.size )
                image.SetData( source.convert( "RGB").tostring() )
                # if the image has an alpha channel, you can set it with this line:
                image.SetAlphaData(source.convert("RGBA").tostring()[3::4])
                return image.ConvertToBitmap() # wx.BitmapFromImage(image)

NumPy

This example shows the creation of a bitmap using a NumPy array as the data source. Note that NumPy uses reversed column-row ordering compared to wxPython, so you'll need to make sure that you generate images using height, width, not width, height coordinates. Also note the use of the 'uint8' data type for image data in rgb format.

        import wx, numpy

        def GetBitmap( self, width=32, height=32, colour = (0,0,0) ):
                array = numpy.zeros( (height, width, 3),'uint8')
                array[:,:,] = colour
                image = wx.EmptyImage(width,height)
                image.SetData( array.tostring())
                return image.ConvertToBitmap() # wx.BitmapFromImage(image)

Slicing

NumPy's slicing notation, for our purposes, works as follows: [ rowStart: rowEnd, columnStart: columnEnd, colourPlanes ] To assign a single colour to the entire image:
- array[:,:] = (r,g,b)
To assign a value to the entire red bit plane:
- array[:,:,0] = 255
To assign a colour to the first row:
- array[ 0 ] = (r,g,b )

Examples

This example shows use of NumPy's (extended) slicing notation to alter an in-memory image. Note that the slice notation allows for standard negative-value index semantics. Also note the accomodating nature of the assignment operation. Assigning a three-tuple to a row sets every value in the row, while assigning a list of three-tuples will set each value individually (and raise an error if the number of items is incorrect).

        import wx, numpy

        def GetBitmap( self, width=32, height=32, colour = (128,128,128), border=5, borderColour=(255,255,255) ):
                # creates a bitmap with a border
                array = numpy.zeros( (height, width, 3), 'uint8')
                array[border:-border,border:-border,:] = colour
                array[:border,:,:] = borderColour
                array[-border:,:,:] = borderColour
                array[:,:border,:] = borderColour
                array[:,-border:,:] = borderColour
                image = wx.EmptyImage(width,height)
                image.SetData( array.tostring())
                return image.ConvertToBitmap()# wx.BitmapFromImage(image)

This example creates a horizontal gradient between two colours. Note that NumPy is automatically converting from float to uint8 data type in the assignment to array, because of the slice indexing. Without the indexing, array would just get overwritten to point to a different numpy array instead of assigning data within the existing one.

        import numpy as np

        def GetBitmap( self, width=640, height=480, leftColour=(255,128,0), rightColour=(64,0,255) ):
                ## Create a horizontal gradient
                array = np.zeros( (height, width, 3),'uint8')
                # alpha is a one dimensional array with a linear gradient from 0.0 to 1.0
                alpha = np.linspace( 0., 1., width )
                # This uses alpha to linearly interpolate between leftColour and rightColour
                colourGradient = np.outer(alpha, leftColour) + np.outer((1.-alpha), rightColour)
                # NumPy's broadcasting rules will assign colourGradient to every row of the destination array
                array[:,:,:] = colourGradient
                image = wx.EmptyImage(width,height)
                image.SetData( array.tostring())
                return image.ConvertToBitmap()# wx.BitmapFromImage(image)

Screen Capture

This example captures the client area of a frame to a wxBitmap object (and from there, a file). Note that there is no error checking done here. Should likely use Ok() or something to check each context. I'm not sure what would happen if this were called before the first painting of the screen, likely the context would have the desktop or something in it. Note: code below only captures the client area of the window. Tagore notes that you can get the whole window with a wxWindowDC and arbitrary parts of the screen with a wxScreenDC.

        def OnToFile( self, event ):
                context = wx.ClientDC( self )
                memory = wx.MemoryDC( )
                x, y = self.ClientSize
                bitmap = wx.EmptyBitmap( x, y, -1 )
                memory.SelectObject( bitmap )
                memory.Blit( 0, 0, x, y, context, 0, 0)
                memory.SelectObject( wx.NullBitmap)
                bitmap.SaveFile( "test.bmp", wx.BITMAP_TYPE_BMP )

A Flexible Screen Capture App

Here's a general screen capture app that first captures the whole primary screen, then captures 4 smaller portions of it. General_ScreenShot_Using_WX_Demo.py

All captures are saved to PNG files. The wx.ScreenDC treats the Desktop screen as a single seamless conglomerate of all the wx.Display(n) areas. Negative coordinate values relative to the primary Desktop screen are allowed and necessary to be used if any extension screens left of or above the primary screen are to be captured.

Other wxPython add-on packages can take screen shots, too, such as ImageMagick and PIL: General_ScreenShot_Using_WX_Demo.py -- Ray Pasco

Write Text to Bitmap

Writing text into a bitmap is done by creating a context then using the context's DrawText method to do the actual drawing. A simple example:

from wxPython.wx import *
def _setupContext( memory, font=None, color=None ):
        if font:
                memory.SetFont( font )
        else:
                memory.SetFont( wxNullFont )
        if color:
                memory.SetTextForeground( color )
def write(  text, bitmap, pos=(0,0), font=None, color=None):
        """Simple write into a bitmap doesn't do any checking."""
        memory = wx.MemoryDC( )
        _setupContext( memory, font, color )
        memory.SelectObject( bitmap )
        try:
                memory.DrawText( text, pos[0],pos[1],)
        finally:
                memory.SelectObject( wxNullBitmap)
        return bitmap

One of the things you might want to use this for is creating "button labels" for wxBitmapButtons. The following utility function provides basic centred-text (potentially multi-line) captions for your bitmaps. It also demonstrates the use of the context's GetTextExtent method to perform centring of the text:

from wxPython.wx import *

MINIMUMFONTSIZE = 4

def writeCaption( text, bitmap, font=None, margins = (2,2), color=None ):
        """Write the given caption (text) into the bitmap
        using the font (or default if not given) with the
        margins given.  Will try to make sure the text fits
        into the bitmap.
        """
        memory = wx.MemoryDC( )
        font = font or memory.GetFont()
        textLines = text.split( '\n' )
        fit = False
        while not fit:
                totalWidth=0
                totalHeight = 0
                setLines = []
                for line in textLines:
                        if line and line[-1] == '\r':
                                line = line[:-1]
                        width, height = extents = memory.GetTextExtent( line )
                        totalWidth = max( totalWidth,  width )
                        totalHeight += height
                        setLines.append( (line, extents))
                if (
                        totalWidth > (bitmap.GetWidth()- 2*margins[0]) or
                        totalHeight > (bitmap.GetHeight()- 2*margins[0])
                ):
                        size = font.GetPointSize()-1
                        if size < MINIMUMFONTSIZE:
                                fit = True # will overdraw!!!
                        else:
                                font.SetPointSize( size )
                                memory.SetFont( font )
                else:
                        fit = True
        if not setLines:
                return bitmap
        centreX, centreY = (bitmap.GetWidth()/2), (bitmap.GetHeight()/2)
        x, y = centreX-(totalWidth/2), centreY-(totalHeight/2)
        memory.SelectObject( bitmap )
        _setupContext( memory, font, color)
        for line, (deltaX, deltaY) in setLines:
                x = centreX - (deltaX/2)
                memory.DrawText( line, x, y,)
                y += deltaY
        memory.SelectObject( wxNullBitmap)
        return bitmap

def _setupContext( memory, font=None, color=None ):
        if font:
                memory.SetFont( font )
        else:
                memory.SetFont( wxNullFont )
        if color:
                memory.SetTextForeground( color )

Creating a Variant of a Bitmap with Transparency

To copy a bitmap with a transparent background and draw some more stuff on top, try the following:

# make a new, empty bitmap
w = oldPix.GetWidth();
h = oldPix.GetHeight();
newPix = wxEmptyBitmap(w, h, oldPix.GetDepth());

# create a DC for drawing to it
mem = wxMemoryDC();
mem.SelectObject(newPix);

# do your custom drawing -- stuff drawn before the old
# bitmap will underlay it, stuff drawn after will overlay it
mem.SetPen(wxMEDIUM_GREY_PEN);
mem.DrawLines(((0, h-1), (0, 0), (w-1, 0)));

# now draw the source bitmap contents
mem.DrawBitmap(oldPix, 0, 0, 1);

# all done -- disconnect the DC
mem.SelectObject(wxNullBitmap);

# now get the transparency set up by assigning a new mask
# that is derived from the current contents of the new
# bitmap -- choose the color you use carefully!
newPix.SetMask(wxMaskColour(newPix, wxBLACK));

Since transparency is a bit tricky here is another example. (Which unlike the above, works for me).

In this example two bitmaps are created. One is completly white (oldPix) the other one completly black (newPix). Then a red box is drawn on the black bitmap. Now the black color in newPix is masked and then newPix is copied over oldPix (the white one). The result is, that only the red box is copied.

from wxPython.wx import *

class Test(wxApp):
    def OnInit(self):
        oldPix  = wx.wxEmptyBitmap(300, 300)
        newPix  = wx.wxEmptyBitmap(300, 300)
        mem = wxMemoryDC()
        mem.SelectObject(oldPix)
        mem.Clear()                       # The images have to be cleared
        mem.SelectObject(newPix)          # because wxEmptyBitmap only
        mem.SetBackground(wxBLACK_BRUSH)  # allocates the space
        mem.Clear()

        # We now have a black and a white image
        # Next we plot a white box in the middle of the black image

        mem.SetPen(wxRED_PEN)
        mem.DrawLines(((100, 200), (100, 100), (200, 100), (200,200), (100,200)))

        mem.SelectObject(oldPix)
        newPix.SetMask(wxMaskColour(newPix, wxBLACK))
        mem.DrawBitmap(newPix, 0, 0, 1)

        oldPix.SaveFile("oldPix.bmp", wx.wxBITMAP_TYPE_BMP)
        newPix.SaveFile("newPix.bmp", wx.wxBITMAP_TYPE_BMP)
        return true


app = Test(0)
app.MainLoop()

This is a complete wxPython app, so just run it and have a look at the created Bitmaps oldPix.bmp and newPix.bmp for the results. -- Nikolai Hlubek

PythonMagick (a 16-bit Imaging Library)

Here is a short but complete PythonMagick/wxPython program. The program uses wxPython GUI, loads an image, displays the image, and does a simple image processing operation (threshold). Various conversions between PythonMagick, PIL, and wxPython images are available on the PythonMagick web site.

See the PythonMagick page for a greater description of PythonMagick.

Note that PythonMagick presently only has a Windows installer.