Contents | Previous | Next

Chapter    5


The Java 2D™ API supports three imaging models

  • The producer/consumer (push) model provided in previous versions of the JDK software.
  • The immediate mode model introduced in the Java™ 2 SDK software release.
  • The pipeline (pull) model compatible with the immediate mode model and that will be fully implemented in the forthcoming Java Advanced Imaging API.

The following table contrasts the features of each of these imaging models.

Push Model
Immediate Mode Image Buffer Model
Pull Model
Major Interfaces/Classes
  • Image
  • ImageProducer
  • ImageConsumer
  • ImageObserver
    (JDK 1.0.x, 1.1.x)
  • BufferedImage
  • Raster
  • BufferedImageOp
  • RasterOp
    (JavaTM 2D API)
  • RenderableImage
  • RenderableImageOp
    (Java 2D API)
  • RenderedOp
  • RenderableOp
  • TiledImage
    (Java Advanced Imaging API)
  • Processing driven by image availability (e.g. over network)
  • Images processed incrementally
  • Simplest programming interface
  • Commonly used model
  • Stores/processes only required data
  • Allows lazy evaluation
  • Requires transfer (but not processing)
  • More complex programming interface
  • Requires memory allocation of complete images
  • Requires processing of complete images
  • More complex programming interface
  • More complex implementation

This chapter focuses on the objects and techniques of the immediate mode imaging model. The immediate mode imaging classes and interfaces of the Java 2D API provide techniques for dealing with pixel mapped images whose data is stored in memory. This API supports accessing image data in a variety of storage formats and manipulating image data through several types of filtering operations.

5.1 Interfaces and Classes

The immediate mode imaging APIs in the Java 2D™ API can be grouped into six categories: interfaces, image data classes, image operation classes, sample model classes, color model classes, and exceptions.

5.1.1 Imaging Interfaces

Describes single-input/single-output operations performed on BufferedImage objects. Implemented by AffineTransformOp, ColorConvertOp, ConvolveOp, LookupOp, and RescaleOp.
Defines single-input/single-output operations performed on Raster objects. Implemented by AffineTransformOp, BandCombineOp, ColorConvertOp, ConvolveOp, LookupOp, and RescaleOp.
Defines a common protocol for objects that contain or can produce image data in the form of Rasters.
Extends: RenderedImage
Defines a common protocol for objects that contain or can produce image data in the form of Rasters which can be modified.
Defines a protocol for objects that want to be notified when the modification state of a WritableRenderedImage changes.

5.1.2 Image Data Classes

Extends: Image
Implements: WriteableRenderedImage
An image with an accessible data buffer. A BufferedImage has a ColorModel and a Raster of image data.
Extends: LookupTable
A LookupTable that contains byte data.
Wraps one or more data arrays holding pixel data. Each data array is called a bank.
Extends: DataBuffer (Final)
A data buffer that stores bytes of data. (Used in Java Advanced Imaging API)
Extends: DataBuffer (Final))
A data buffer that stores integer data.(Used in Java Advanced Imaging API)
Extends: DataBuffer (Final)
A data buffer that stores short data.(Used in Java Advanced Imaging API)
Extends: DataBuffer (Final)
A data buffer that stores unsigned short data.
A matrix that describes how an input pixel and its surrounding pixels affect the value of an output pixel in a ConvolveOp filtering operation.
Extends: Object
A table that maps values from single-banded pixel data to color values.
A rectangular array of pixels from which you can retrieve image data. A Raster contains a DataBuffer and a SampleModel.
Extends: LookupTable
A lookup table that contains short data.
Extends: Raster
A Raster that you can modify.

5.1.3 Image Operation Classes

Implements: BufferedImageOp, RasterOp
A class that defines an affine transform to perform a linear mapping from 2D coordinates in a source Image or Raster to 2D coordinates in the destination image or Raster. This class can perform either bilinear or nearest neighbor affine transform operations.
Implements: RasterOp
Using a specified matrix, this operation performs an arbitrary linear combination of bands in a Raster.
Extends: ImageFilter
An ImageFilter that provides a simple means of using a BufferedImageOp (a single-source/single-destination image operator) to filter a BufferedImage or Raster.
Implements: BufferedImageOp, RasterOp
Performs a pixel-by-pixel color conversion of the data in the source image.
Implements: BufferedImageOp, RasterOp
Uses a Kernel to perform a convolution on the source image. A convolution is a spatial operation where the pixels surrounding the input pixel are multiplied by a kernel value to generate the value of the output pixel. The Kernel mathematically defines the relationship between the pixels in the immediate neighborhood of the input pixel and the output pixel.
Implements: BufferedImageOp, RasterOp
Performs a lookup operation from the source to the destination. For Rasters, the lookup operates on sample values. For BufferedImages, the lookup operates on color and alpha components.
Implements: BufferedImageOp, RasterOp
Performs a pixel-by-pixel rescaling of the data in the source image by multiplying each pixel value by a scale factor and then adding an offset.

5.1.4 Sample Model Classes

Extends: ComponentSampleModel (Final)
Provides access to image data stored with like samples stored as bands in separate banks of a DataBuffer. A pixel consists of one sample from each band.
Extends: SampleModel
Provides access to image data stored with each sample of a pixel residing in a separate element of a DataBuffer. Different types of pixel interleaving are supported.
Extends: SampleModel
Provides access to image data stored with multiple one-sample pixels packed into one element of a DataBuffer.
Extends: ComponentSampleModel
Provides access to image data stored with the sample data for each pixel in adjacent elements of the data array, and all elements in a single bank of a DataBuffer.
An abstract class that defines a mechanism for extracting sample data from an image without knowing how the underlying data is stored in a DataBuffer.
Extends: SampleModel
Provides access to image data stored with all the samples belonging to an individual pixel packed into one element of a DataBuffer.

5.1.5 Color Model Classes

Implements: Transparency
JDK1.1 class. An abstract class that defines methods for translating from image pixel values to color components such as red, green, and blue.
Extends: ColorModel
A ColorModel that can handle an arbitrary ColorSpace and an array of color components to match the ColorSpace. This class can be used to represent most color models on most types of GraphicsDevices.
Extends: PackedColorModel
JDK1.1 class. A ColorModel that represents pixel values that have RGB color components embedded directly in the bits of the pixel. This color model is similar to an X11 TrueColor visual. The default RGB ColorModel returned by ColorModel.getRGBdefault is a DirectColorModel.
Extends: ColorModel
JDK1.1 class. A ColorModel that represents pixel values that are indices into a fixed color map in the sRGB ColorSpace.
Extends: ColorModel
An abstract ColorModel that represents pixel values that have color components embedded directly in the bits of a pixel. DirectColorModel extends PackedColorModel to support pixels that contain RGB color components.

5.1.6 Exception Classes

Extends: RuntimeException
Thrown if one of the BufferedImageOp or RasterOp filter methods can’t process the image.
Extends: RuntimeException
Thrown if there is invalid layout information in the Raster.

5.2 Immediate Mode Imaging Concepts

The immediate mode imaging model supports fixed-resolution images stored in memory. The model also supports filtering operations on image data. A number of classes and interfaces are used in this model.

The following context describes this graphic.

Figure 5-1 BufferedImage and supporting classes

As shown in Figure 5-1, BufferedImage provides general image management. A BufferedImage can be created directly in memory and used to hold and manipulate image data retrieved from a file or URL. A BufferedImage can be displayed using any Graphics2D object for a screen device, or rendered to any other destination using appropriate Graphics2D context. A BufferedImage object contains two other objects: a Raster and a ColorModel.

The Raster class provides image data management. It represents the rectangular coordinates of the image, maintains image data in memory, and provides a mechanism for creating multiple subimages from a single image data buffer. It also provides methods for accessing specific pixels within an image. A Raster object contains two other objects, a DataBuffer and a SampleModel.

The DataBuffer class holds pixel data in memory.

The SampleModel class interprets data in the buffer and provides it as individual pixels or rectangular ranges of pixels.

The ColorModel class provides a color interpretation of pixel data provided by the image’s sample model.

The image package provides additional classes that define filtering operations on BufferedImage and Raster objects. Each image processing operation is embodied in a class that implements the BufferedImageOp interface, the RasterOp interface, or both interfaces. The operation class defines filter methods that performs the actual image manipulation.

Figure 5-2 illustrates the basic model for Java 2D™ API image processing:

Flow diagram shows that the source image flows through an image-processing operation before becoming the destination image.

Figure 5-2 Image Processing Model

The operations supported include:

  • Affine transformation
  • Amplitude scaling
  • Lookup-table modification
  • Linear combination of bands
  • Color conversion
  • Convolution

Note that if you’re interested just in displaying and manipulating images, you only need to understand the BufferedImage class and the filtering operation classes. On the other hand, if you’re planning to write filters or otherwise directly access image data, you’ll need to understand the classes associated with BufferedImage.

5.2.1 Terminology

Here are some terms used throughout the following discussions:

Data Elements: primitive types used as units of storage of image data. Data elements are individual members of a DataBuffer array. The layout of elements in the data buffer is independent of the interpretation of the data as pixels by an image’s SampleModel.

Samples: distinct members of the pixels of an image. A SampleModel provides a mechanism for converting elements in the DataBuffer to pixels and their samples. The samples of a pixel may represent primary values in a particular color model. For example, a pixel in an RGB color model consists of three samples: red, green, and blue.

Components: values of pixels independent of color interpretation. The distinction between component and sample is useful with IndexColorModel, where pixel components are indexes into the LookupTable.

Band: the set of all samples of one type in an image, such as all red samples or all green samples. Pixel data can be stored in a number of ways, the two supported in the Java 2D API being banded and pixel interleaved. Banded storage organizes image data by bands, and a pixel is made up of sample data from the same position in each band. Pixel interleaved storage organizes image data by pixels, with a single array containing all pixels, and bands consisting of the set of samples at the same index position in each pixel.

Primaries: distinct members of a color value in a specific color model; for example the RGB model forms color values from the primaries red, green, and blue.

5.3 Using BufferedImages

The BufferedImage class is the main class supporting the immediate imaging mode. It manages an image in memory, providing ways to store pixel data, interpret pixel data, and to render the pixel data to a Graphics or Graphics2D context.

5.3.1 Creating a BufferedImage

To create a BufferedImage, call the Component.createImage method; this returns a BufferedImage whose drawing characteristics match those of the component used to create it—the created image is opaque, has the foreground and background colors of the Component, and you can’t adjust the transparency of the image. You could use this technique when you want to do double buffered drawing for animation in a component; the discussion “Drawing in an Offscreen Buffer” on page 79 gives more details.

    public Graphics2D createDemoGraphics2D(Graphics g) { 
        Graphics2D g2 = null; 
        int width = getSize().width;  
        int height = getSize().height;  
        if (offImg == null || offImg.getWidth() != width || 
                        offImg.getHeight() != height) { 
            offImg = (BufferedImage) createImage(width, height); 
        if (offImg != null) { 
            g2 = offImg.createGraphics(); 
        // .. clear canvas .. 
        g2.clearRect(0, 0, width, height); 
        return g2; 

You can also create a blank BufferedImage in memory using one of several constructor methods provided.

5.3.2 Drawing in an Offscreen Buffer

The BufferedImage class can be used to prepare graphic elements offscreen then copy them to the screen. This technique is especially useful when a graphic is complex or used repeatedly. For example, if you want to display a complicated shape several times, you could draw it once into an offscreen buffer and then copy it to different locations in the window. By drawing the shape once and copying it, you can display the graphics more quickly.

The java.awt package facilitates the use of offscreen buffers by letting you draw to an Image object the same way that you draw to a window. All of the Java 2D™ API rendering features can be used when drawing to offscreen images.

Offscreen buffers are often used for animation. For example, you could use an offscreen buffer to draw an object once and then move it around in a window. Similarly, you could use an offscreen buffer to provide feedback as a user moves a graphic using the mouse. Instead of redrawing the graphic at every mouse location, you could draw the graphic once to an offscreen buffer, and then copy it to the mouse location as the user drags the mouse.1

The following context describes this graphic.

Figure 5-3 Using an Offscreen Buffer

Figure 5-3 demonstrates how a program can draw to an offscreen image and then copy that image into a window multiple times. The last time the image is copied, it is transformed. Note that transforming the image instead of redrawing it with the transformation might produce unsatisfactory results. Creating an Offscreen Buffer

The simplest way to create an image that you can use as an offscreen buffer is to use the Component.createImage method.

By creating an image whose color space, depth, and pixel layout exactly match the window into which you are drawing, the image can be efficiently blitted to a graphics device. This allows drawImage to do its job quickly.

You can also construct a BufferedImage object directly to use as an offscreen buffer. This is useful when you need control over the offscreen image’s type or transparency.

BufferedImage supports several predefined image types:


A BufferedImage object can contain an alpha channel. In Figure 5-3, an alpha channel is used to distinguish painted and unpainted areas, allowing an irregular shape to appear over graphics that have already been painted (in this case, a shaded rectangle). In other cases, you might use alpha channel to blend the colors of the new image into those in the existing image.

Note: unless you need alpha image data for transparency, as with the irregularly shaped images shown in Figure 5-2, you should avoid creating an off-screen buffer with alpha. Using alpha where it’s unnecessary slows rendering performance.

GraphicsConfiguration provides convenience methods that automatically create buffered images in a format compatible with your configuration. You can also query the graphics configuration associated with the graphics device on which the window resides to get the information you need to construct a compatible BufferedImage object. Drawing in an Offscreen Buffer

To draw in a buffered image, you call its BufferedImage.createGraphics method, which returns a Graphics2D object. With this object, you can call all of the Graphics2D methods to draw graphics primitives, place text, and render other images in the image. This drawing technique supports dithering and other enhancements provided by the 2D imaging package. The following code illustrates the use of offscreen buffering:

    public void update(Graphics g){ 
        Graphics2D g2 = (Graphics2D)g; 
            Dimension dim = getSize(); 
            int w = dim.width; 
            int h = dim.height; 
            area = new Rectangle(dim); 
            bi = (BufferedImage)createImage(w, h); 
            big = bi.createGraphics(); 
            rect.setLocation(w/2-50, h/2-25); 
            big.setStroke(new BasicStroke(8.0f)); 
            firstTime = false; 
        // Clears the rectangle that was previously drawn. 
        big.clearRect(0, 0, area.width, area.height); 
        // Draws and fills the newly positioned rectangle to the buffer. 
        // Draws the buffered image to the screen. 
        g2.drawImage(bi, 0, 0, this); 

5.3.3 Manipulating BufferedImage Data Directly

In addition to drawing directly in a BufferedImage, you can directly access and manipulate the image’s pixel data in a couple of ways. These are useful if you’re implementing the BufferedImageOp filtering interface, as described in “Image Processing and Enhancement” on page 84.

You can use the BufferedImage.setRGB methods to directly set the value of a pixel or a pixel array to a specific RGB value. Note that no dithering is performed when you modify pixels directly. You can also manipulate pixel data by manipulating a WritableRaster object associated with a BufferedImage (see“Managing and Manipulating Rasters” on page 80).

5.3.4 Filtering a BufferedImage

You can apply a filtering operation to a BufferedImage using an object that implements BufferedImageOp interface. Filtering and the classes that provide this filtering interface are discussed in “Image Processing and Enhancement” on page 84.

5.3.5 Rendering a BufferedImage

To render a buffered image into a specific context, call one of the drawImage method of the context’s Graphics object. For example, when rendering within a Component.paint method, you call drawImage on the graphics object passed to the method.

    public void paint(Graphics g) { 
        if (getSize().width <= 0 || getSize().height <= 0) 
        Graphics2D g2 = (Graphics2D) g; 
        if (offImg != null && isShowing())  { 
            g2.drawImage(offImg, 0, 0, this); 

5.4 Managing and Manipulating Rasters

A BufferedImage object uses a Raster to manage its rectangular array of pixel data. The Raster class defines fields for the image’s coordinate system—width, height, and origin. A Raster object itself uses two objects to manage the pixel data, a DataBuffer and a SampleModel. The DataBuffer is the object that stores pixel data for the raster (as described on page 82), and the SampleModel provides the interpretation of pixel data from the DataBuffer (as described on page 82).

5.4.1 Creating a Raster

In most cases, you don’t need to create a Raster directly, since one is supplied with any BufferedImage that you create in memory. However, one of the BufferedImage constructor methods allows you to create a Raster by passing in a WritableRaster.

The Raster class provides a number of static factory methods for creating Rasters with the DataBuffers and SampleModels you specify. You can use these factories when implementing RasterOp filtering classes.

5.4.2 Parent and Child Rasters

The Raster class incorporates the concept of parent and child rasters. This can improve storage efficiency by allowing you to construct any number of buffered images from the same parent. The parent and its children all refer to the same data buffer, and each child has a specific offset and bounds to identify its image location in the buffer. A child identifies its ownership through its getParent method.

To create a subraster, you use the Raster.createSubRaster method.When you create a subraster, you identify the area of its parent that it covers and its offset from the parent’s origin.

5.4.3 Operations on a Raster

The Raster class defines a number of ways to access pixels and pixel data. These are useful when you’re implementing the RasterOp interface, which provides raster-level filtering and manipulation of image data, or when implementing any method that needs to perform low-level pixel manipulation.

The Raster.getPixel methods let you get an individual pixel, which is returned as individual samples in an array. The Raster.getDataElements methods return a specified run of uninterpreted image data from the DataBuffer. The Raster.getSample method returns samples of an individual pixel. The getSamples method returns a band for a particular region of an image.

In addition to these methods, you can also access the data buffer and the sample model through instance variables of the Raster class. These objects provide additional ways to access and interpret the Raster’s pixel data.

5.4.4 The WritableRaster Subclass

The WritableRaster subclass provides methods for setting pixel data and samples. The Raster associated with a BufferedImage is actually a WritableRaster, thus providing full access to manipulate its pixel data.

5.5 Image Data and DataBuffers

The DataBuffer belonging to a Raster represents an array of image data. When you create a Raster directly or through the BufferedImage constructors, you specify a width and height in pixels, along with a SampleModel for the image data. This information is used to create a DataBuffer of the appropriate data type and size.

There are three subclasses of DataBuffer, each representing a different type of data element:

  • DataBufferByte (represents 8-bit values)
  • DataBufferInt (represents 32-bit values)
  • DataBufferShort (represents 16-bit values)
  • DataBufferUShort (represents unsigned short values)

As defined earlier, elements are the discrete members of the array of the data buffer, and components or samples are the discrete values that together make up a pixel. There can be various mappings between a particular type of element in a DataBuffer and a particular type of pixel represented by a SampleModel. It is the responsibility of the various SampleModel subclasses to implement that mapping and provide a way to get specific pixels from a specific DataBuffer.

DataBuffer constructors provide ways to create buffers of a specific size and a specific number of banks.

While you can access image data in a DataBuffer directly, it’s generally easier and more convenient to do so through the methods of the Raster and WritableRaster classes.

5.6 Extracting Pixel Data from a SampleModel

The abstract SampleModel class defines methods for extracting samples of an image without knowing how the underlying data is stored. The class provides fields for tracking the height and width of the image data in the associated DataBuffer, and for describing the number of bands and the data type of that buffer. SampleModel methods provide image data as a collection of pixels, with each pixel consisting of a number of samples or components.

The java.awt.image package provides five types of sample models:

  • ComponentSampleModel—used to extract pixels from images that store sample data in separate data array elements in one bank of a DataBuffer.
  • BandedSampleModel—used to extract pixels from images that store each sample in a separate data element with bands stored in a sequence of data elements
  • PixelInterleavedSampleModel—used to extract pixels from images that store each sample in a separate data element with pixels stored in a sequence of data elements.
  • MultiPixelPackedSampleModel—used to extract pixels from single banded images that store multiple one-sample pixels in one data element.
  • SinglePixelPackedSampleModel—used to extract samples from images that store sample data for a single pixel in one data array element in the first bank of a DataBuffer.

Pixel data presented by the SampleModel may or may not correlate directly to a color data representation of a particular color model, depending on the data source. For example, in photographic image data, the samples may represent RGB data. In image data from a medical imaging device, samples can represent different types of data such as temperature or bone density.

There are three categories of methods for accessing image data. The getPixel methods return a whole pixel as an array, with one entry for each sample. The getDataElement methods provide access to the raw, uninterpreted data stored in the DataBuffer. The getSample methods provide access to pixel components for a specific band.

5.7 ColorModels and Color Data

In addition to the Raster object for managing image data, the BufferedImage class includes a ColorModel for interpreting that data as color pixel values. The abstract ColorModel class defines methods for turning an image’s pixel data into a color value in its associated ColorSpace.

The java.awt.image package provides four types of color models:

  • PackedColorModel—An abstract ColorModel that represents pixel values that have color components embedded directly in the bits of an integer pixel. A DirectColorModel is a subclass of PackedColorModel.
  • DirectColorModel—a ColorModel that represents pixel values that have RGB color components embedded directly in the bits of the pixel itself. DirectColorModel model is similar to an X11 TrueColor visual.
  • ComponentColorModel—a ColorModel that can handle an arbitrary ColorSpace and an array of color components to match the ColorSpace.
  • IndexColorModel—a ColorModel that represents pixel values that are indices into a fixed color map in the sRGB color space.

ComponentColorModel and PackedColorModel are new in the Java™ 2 SDK software release.

Based on data in the DataBuffer, the SampleModel provides the ColorModel with a pixel, which the ColorModel then interprets as a color.

5.7.1 Lookup Table

A lookup table contains data for one or more channels or image components; for example, separate arrays for R, G, and B. The java.awt.image package defines two types of lookup tables that extend the abstract LookupTable class, one that contains byte data and one that contains short data (ByteLookupTable and ShortLookupData).

5.8 Image Processing and Enhancement

The image package provides a pair of interfaces that define operations on BufferedImage and Raster objects: BufferedImageOp and RasterOp.

The classes that implement these interfaces include AffineTransformOp, BandCombineOp, ColorConvertOp, ConvolveOp, LookupOp, RescaleOp. These classes can be used to geometrically transform, blur, sharpen, enhance contrast, threshold, and color correct images.

Figure 5-4 illustrates edge detection and enhancement, an operation that emphasizes sharp changes in intensity within an image. Edge detection is commonly used in medical imaging and mapping applications. Edge detection is used to increase the contrast between adjacent structures in an image, allowing the viewer to discriminate greater detail.

The previous context describes this graphic.

Figure 5-4 Edge detection and enhancement

The following code illustrates edge detection:

float[] elements = { 0.0f, -1.0f, 0.0f, 
                    -1.0f, 4.f, -1.0f, 
                    0.0f, -1.0f, 0.0f}; 
BufferedImage bimg = new BufferedImage(bw,bh,BufferedImage.TYPE_INT_RGB); 
Kernel kernel = new Kernel(3, 3, elements); 
ConvolveOp cop = new ConvolveOp(kernel, ConvolveOp.EDGE_NO_OP,                                             null); 

Figure 5-5 demonstrates lookup table manipulation. A lookup operation can be used to alter individual components of a pixel.

The previous context describes this graphic.

Figure 5-5 Lookup-table Manipulation

The following code demonstrates Lookup-table manipulation:

        byte reverse[] = new byte[256]; 
   for (int j=0; j<200; j++){  
        ByteLookupTable blut=new ByteLookupTable(0, reverse);  
        LookupOp lop = new LookupOp(blut, null);  

Figure 5-6 illustrates rescaling. Rescaling can increase or decrease the intensity of all points. Rescaling can be used to increase the dynamic range of an otherwise neutral image, bringing out detail in a region that appears neutral or flat.

The previous context describes this graphic.

Figure 5-6 Rescaling

The following code snippet illustrates rescaling:

        RescaleOp rop = new RescaleOp(1.5f, 1.0f, null); 

5.8.1 Using an Image Processing Operation

Convolution is the process that underlies most spatial filtering algorithms. Convolution is the process of weighting or averaging the value of each pixel in an image with the values of neighboring pixels.This allows each output pixel to be affected by the immediate neighborhood in a way that can be mathematically specified with a kernel. Figure 5-7 illustrates Convolution.

The previous context describes this graphic.

Figure 5-7 Blurring with Convolution

The following code fragment illustrates how to use one of the image processing classes, ConvolveOp. In this example, each pixel in the source image is averaged equally with the eight pixels that surround it.

float weight = 1.0f/9.0f;float[] elements = new float[9]; // create 2D array// fill the array with nine equal elements 
for (i = 0; i < 9; i++) {   elements[i] = weight;}// use the array of elements as argument to create a Kernelprivate Kernel myKernel = new Kernel(3, 3, elements);public ConvolveOp simpleBlur = new ConvolveOp(myKernel); 
// sourceImage and destImage are instances of BufferedImagesimpleBlur.filter(sourceImage, destImage) // blur the image 

The variable simpleBlur contains a new instance of ConvolveOp that implements a blur operation on a BufferedImage or a Raster. Suppose that sourceImage and destImage are two instances of BufferedImage. When you call filter, the core method of the ConvolveOp class, it sets the value of each pixel in the destination image by averaging the corresponding pixel in the source image with the eight pixels that surround it.

The convolution kernel in this example could be represented by the following matrix, with elements specified to four significant figures:

A 3 by 3 matrix with all values being 0.1111.

When an image is convolved, the value of each pixel in the destination image is calculated by using the kernel as a set of weights to average the pixel’s value with the values of surrounding pixels. This operation is performed on each channel of the image.

The following formula shows how the weights in the kernel are associated with the pixels in the source image when the convolution is performed. Each value in the kernel is tied to a spatial position in the image.

A 3 by 3 matrix. The first row values are i minus 1, j minus 1, i, j minus 1, and i plus 1, j minus1. The second row values are i minus 1, j, i, j, and i plus1, j. The third row values are i minus 1, j plus 1, i, j plus 1, and i plus 1, j plus 1.

The value of a destination pixel is the sum of the products of the weights in the kernel multiplied by the value of the corresponding source pixel. For many simple operations, the kernel is a matrix that is square and symmetric, and the sum of its weights adds up to one.2

The convolution kernel in this example is relatively simple. It weights each pixel from the source image equally. By choosing a kernel that weights the source image at a higher or lower level, a program can increase or decrease the intensity of the destination image. The Kernel object, which is set in the ConvolveOp constructor, determines the type of filtering that is performed. By setting other values, you can perform other types of convolutions, including blurring (such as Gaussian blur, radial blur, and motion blur), sharpening, and smoothing operations. Figure 5-8 illustrates sharpening using Convolution.

The previous context describes this graphic.

Figure 5-8 Sharpening with Convolution

The following code snippet illustrates sharpening with Convolution:

float[] elements = { 0.0f, -1.0f, 0.0f, 
                    -1.0f,  5.f, -1.0f, 
                     0.0f, -1.0f,  0.0f}; 
Kernel kernel = new Kernel(3,3,elements); 
ConvolveOp cop = new ConvolveOp(kernel, ConvolveOp.EDGE_NO_OP,                                                                                                          null); 

1It is up to the programmer to “erase” the previous version of the image before making a new copy at a new location. This can be done by redrawing the background or copying the background from another offscreen buffer.
2If the sum of the weights in the matrix is one, the intensity of the destination image is unchanged from the source.


Contents | Previous | Next

Copyright © 1993, 2014, Oracle and/or its affiliates. All rights reserved.