Enhance Java GUIs with Windows Icons

• By Jeff Friesen
• Apr 4, 2008

Like this article? We recommend
Filthy Rich Clients: Developing Animated and Graphical Effects for Desktop Java™ Applications

Many years ago, Microsoft introduced the Icon Resource Format to specify the internal structure of icons—the Windows operating systems use icons to represent programs and other documents visually. Although Windows icons are often embedded in .EXE and .DLL files, they’re also stored in files with the .ICO extension.

The widespread availability of .ICO files is a good reason to think about using Windows icons with buttons and other GUI components to enhance Java GUIs. Because Java provides no direct support for reading Windows icons from .ICO files, this article introduces a Java library that accomplishes this task.

First we’ll take a tour of Microsoft’s Icon Resource Format. This tour provides insight into how the library works, which is helpful if you ever need to extend the library. After introducing the library, we’ll look at a Swing-based Windows icon viewer application that demonstrates the library’s usefulness.

Tour the Icon Resource Format

Microsoft’s Icon Resource Format specifies the format of a Windows icon resource. As Figure 1 illustrates, the Icon Resource Format organizes the resource into a header, a directory with one or more entries, and one or more images—unlike .ICO files, which adhere to this organization, I’ve encountered .EXE files that don’t store the header or directory portion of the format.

Figure 1 To accommodate different screen resolutions, a Windows icon resource can store multiple images.

The header is a six-byte data structure that begins with a two-byte reserved field, which should contain 0. This field is followed by another two-byte field, which must contain 1 to identify the resource as an icon resource. A third two-byte field completes the header by identifying the number of entries in the directory.

Directory of Images

Because an icon resource can store multiple images, immediately following the header is an image directory. Each directory entry describes one image in terms of its width and height, number of colors, number of color planes, number of bits per pixel, size, and location. The entry is conveniently described via this C structure:

typedef struct
{
  BYTE bWidth;     // Image width (in pixels)
  BYTE bHeight;    // Image height (in pixels)
  BYTE bColorCount;  // Number of image colors (0 if wBitCount is 8 or more)
  BYTE bReserved;   // Reserved (must be 0)
  WORD wPlanes;    // Number of color planes (typically 1)
  WORD wBitCount;   // Number of bits per pixel
  DWORD dwBytesInImage; // Number of bytes making up the image
  DWORD dwImageOffset; // Offset from start of header to the image
}
ICONDIRENTRY;

The bWidth and bHeight members express the image’s width and height dimensions. Although these members can record dimensions from 1×1 to 255×255 (including non-square dimensions such as 48×24), dimensions such as 16×16 and 32×32 are more common because various Windows shells support them.

If bWidth and bHeight contain zeroes, the dimensions must be read from the image data—this is true for those Windows Vista icon images whose dimensions are 256×256. Although the image data could specify higher dimensions such as 1024×768, most (if not all) of the Windows shells don’t support such images.

The bColorCount member records the number of colors. This value is usually two to the power of the wBitCount member’s value. If the value of wBitCount is 8 or higher, the number of colors exceeds 255, bColorCount contains 0, and the number of colors must be read from the image data.

The wPlanes and wBitCount members record information for determining the number of colors (by multiplying their values). Although wPlanes is supposed to be set to 1, some Windows icon resource entries store 0 in this member. In some cases, 0 is also stored in wBitCount.

Finally, the dwBytesInImage and dwImageOffset members record information that’s needed for reading the image data. The first member records the size (in bytes) of the image data, and the second member records the starting location of the image data (relative to the beginning of the header).

Image Data

The directory is followed by a sequence of images, in which each image is stored in one of two formats:

• BITMAPINFOHEADER format
• Portable Network Graphics (PNG)

Let’s look at each format in detail.

BITMAPINFOHEADER Format

This older format expresses an image as a BITMAPINFOHEADER structure followed by an array of RGBQUAD structures, followed by the actual image bits (which are often expressed using XOR and AND bitmaps):

BITMAPINFOHEADER icHeader;  // Device Independent Bitmap (DIB) header
RGBQUAD     icColors []; // Color table
BYTE       icXOR [];  // DIB bits for XOR bitmap
BYTE       icAND [];  // DIB bits for monochrome AND bitmap

Microsoft’s BITMAPINFOHEADER structure, which appears below, provides information necessary for reading an image. Of this structure’s various members, only biSize, biWidth, biHeight, biPlanes, and biBitCount are significant for reading the image data—members other than these and biSizeImage are typically set to 0.

typedef struct
{
  DWORD biSize;
  LONG biWidth;
  LONG biHeight;
  WORD biPlanes;
  WORD biBitCount
  DWORD biCompression;
  DWORD biSizeImage;
  LONG biXPelsPerMeter;
  LONG biYPelsPerMeter;
  DWORD biClrUsed;
  DWORD biClrImportant;
}
BITMAPINFOHEADER;

The biSize member stores the structure’s size, which happens to be 40. A program reading a Windows icon resource checks the first four bytes of image data to see whether they consist of 40 followed by three zeroes. (Remember that this is little-endian byte order.) If this is the case, the program can assume that it has found a BITMAPINFOHEADER structure.

The biWidth and biHeight members store the image’s width and twice its height, respectively. If a directory entry’s width and height are set to 0, these members are accessed to determine the width and height (divided by two). biHeight originally was mandated to contain twice the image height to account for both the XOR and AND bitmaps.

The biPlanes and biBitCount members record information for determining the maximum number of colors used by the image. If a directory entry’s number of colors is 0, these members are accessed to calculate the number of colors. The calculation is expressed as two to the power of the result of multiplying these members’ values by each other.

If the number of colors is 256 or less, the BITMAPINFOHEADER structure is followed by an array of RGBQUAD structures. The number of colors determines the number of entries in this array. For example, there are 2 entries for a 2-color image, 16 entries for a 16-color image, and 256 entries for a 256-color image. Here’s the RGBQUAD structure:

typedef struct
{
  BYTE  rgbBlue;
  BYTE  rgbGreen;
  BYTE  rgbRed;
  BYTE  rgbReserved;
}
RGBQUAD;

If the number of colors exceeds 256, the RGBQUAD array is not present. Pixel values directly describe colors, instead of serving as indexes into this array. For example, if biBitCount contains 32 (24-bit color and an 8-bit alpha channel), each pixel value’s four bytes respectively provide (from low address to high address) the blue, green, red, and alpha color components.

The number of colors determines the manner in which an image is stored. If this value is 256 or less, the image is stored as an XOR bitmap followed by an AND bitmap, in which each bitmap has biWidth by biHeight/2 dimensions. These two bitmaps are used to display images with transparent areas on the screen, as follows:

1. The AND bitmap, a matrix of single-bit values, is first applied to preserve the screen background surrounding the image, and erase the area where the image pixels appear. Existing screen pixels are preserved by ANDing them with the AND bitmap’s corresponding 1 bits; existing screen pixels are erased (made black) by ANDing them with the AND bitmap’s corresponding 0 bits.
2. Next, the XOR bitmap, a matrix of color indexes/values, is applied to display image pixels without affecting the screen background. This is accomplished by XORing black bitmap pixels with existing screen pixels. The black screen pixels (created in the previous step) are XORed with the corresponding bitmap pixels to merge the bitmap’s image with the screen.

If the number of colors describes a 32-bit image, the XOR and AND bitmaps are missing. Instead, a single bitmap with an eight-bit alpha channel is stored. The advantage of the alpha channel over the traditional XOR and AND bitmaps is that the alpha channel makes anti-aliasing possible; jagged edges found in non-horizontal and non-vertical lines (and arcs) are minimized by using various levels of translucency.

When reading an XOR, AND, or 32-bit image bitmap, it’s important to keep in mind that the bitmap is stored upside-down. In other words, the first stored row should appear at the bottom of a displayed image. Another item to remember is that each row of pixel values must be a multiple of four bytes. Zero pad bytes are stored at the end of a row to make sure that the row’s byte length is exactly divisible by four.

Portable Network Graphics (PNG) Format

Some icon resources store image data using the Portable Network Graphics (PNG) format. This format makes it possible to store compressed icon images. Compression is necessary because large images require lots of memory; for example, a single uncompressed 256×256 32-bit image requires 256 kilobytes of storage.