For those methods that use a primitive array pixel representation of type transferType, the array length is the same as the number of color and alpha samples. Color samples are stored first in the array followed by the alpha sample, if present. The order of the color samples is specified by the ColorSpace. Typically, this order reflects the name of the color space type. For example, for TYPE_RGB, index 0 corresponds to red, index 1 to green, and index 2 to blue.

The translation from pixel sample values to color/alpha components for display or processing purposes is based on a one-to-one correspondence of samples to components. Depending on the transfer type used to create an instance of ComponentColorModel, the pixel sample values represented by that instance may be signed or unsigned and may be of integral type or float or double (see below for details). The translation from sample values to normalized color/alpha components must follow certain rules. For float and double samples, the translation is an identity, i.e. normalized component values are equal to the corresponding sample values. For integral samples, the translation should be only a simple scale and offset, where the scale and offset constants may be different for each component. The result of applying the scale and offset constants is a set of color/alpha component values, which are guaranteed to fall within a certain range. Typically, the range for a color component will be the range defined by the getMinValue and getMaxValue methods of the ColorSpace class. The range for an alpha component should be 0.0 to 1.0.

Instances of ComponentColorModel created with transfer types DataBuffer.TYPE_BYTE, DataBuffer.TYPE_USHORT, and DataBuffer.TYPE_INT have pixel sample values which are treated as unsigned integral values. The number of bits in a color or alpha sample of a pixel value might not be the same as the number of bits for the corresponding color or alpha sample passed to the ComponentColorModel(ColorSpace, int[], boolean, boolean, int, int) constructor. In that case, this class assumes that the least significant n bits of a sample value hold the component value, where n is the number of significant bits for the component passed to the constructor. It also assumes that any higher-order bits in a sample value are zero. Thus, sample values range from 0 to 2ⁿ - 1. This class maps these sample values to normalized color component values such that 0 maps to the value obtained from the ColorSpace's getMinValue method for each component and 2ⁿ - 1 maps to the value obtained from getMaxValue. To create a ComponentColorModel with a different color sample mapping requires subclassing this class and overriding the getNormalizedComponents(Object, float[], int) method. The mapping for an alpha sample always maps 0 to 0.0 and 2ⁿ - 1 to 1.0.

For instances with unsigned sample values, the unnormalized color/alpha component representation is only supported if two conditions hold. First, sample value 0 must map to normalized component value 0.0 and sample value 2ⁿ - 1 to 1.0. Second the min/max range of all color components of the ColorSpace must be 0.0 to 1.0. In this case, the component representation is the n least significant bits of the corresponding sample. Thus each component is an unsigned integral value between 0 and 2ⁿ - 1, where n is the number of significant bits for a particular component. If these conditions are not met, any method taking an unnormalized component argument will throw an IllegalArgumentException.

Instances of ComponentColorModel created with transfer types DataBuffer.TYPE_SHORT, DataBuffer.TYPE_FLOAT, and DataBuffer.TYPE_DOUBLE have pixel sample values which are treated as signed short, float, or double values. Such instances do not support the unnormalized color/alpha component representation, so any methods taking such a representation as an argument will throw an IllegalArgumentException when called on one of these instances. The normalized component values of instances of this class have a range which depends on the transfer type as follows: for float samples, the full range of the float data type; for double samples, the full range of the float data type (resulting from casting double to float); for short samples, from approximately -maxVal to +maxVal, where maxVal is the per component maximum value for the ColorSpace (-32767 maps to -maxVal, 0 maps to 0.0, and 32767 maps to +maxVal). A subclass may override the scaling for short sample values to normalized component values by overriding the getNormalizedComponents(Object, float[], int) method. For float and double samples, the normalized component values are taken to be equal to the corresponding sample values, and subclasses should not attempt to add any non-identity scaling for these transfer types.

Instances of ComponentColorModel created with transfer types DataBuffer.TYPE_SHORT, DataBuffer.TYPE_FLOAT, and DataBuffer.TYPE_DOUBLE use all the bits of all sample values. Thus all color/alpha components have 16 bits when using DataBuffer.TYPE_SHORT, 32 bits when using DataBuffer.TYPE_FLOAT, and 64 bits when using DataBuffer.TYPE_DOUBLE. When the ComponentColorModel(ColorSpace, int[], boolean, boolean, int, int) form of constructor is used with one of these transfer types, the bits array argument is ignored.

It is possible to have color/alpha sample values which cannot be reasonably interpreted as component values for rendering. This can happen when ComponentColorModel is subclassed to override the mapping of unsigned sample values to normalized color component values or when signed sample values outside a certain range are used. (As an example, specifying an alpha component as a signed short value outside the range 0 to 32767, normalized range 0.0 to 1.0, can lead to unexpected results.) It is the responsibility of applications to appropriately scale pixel data before rendering such that color components fall within the normalized range of the ColorSpace (obtained using the getMinValue and getMaxValue methods of the ColorSpace class) and the alpha component is between 0.0 and 1.0. If color or alpha component values fall outside these ranges, rendering results are indeterminate.

Methods that use a single int pixel representation throw an IllegalArgumentException, unless the number of components for the ComponentColorModel is one and the component value is unsigned -- in other words, a single color component using a transfer type of DataBuffer.TYPE_BYTE, DataBuffer.TYPE_USHORT, or DataBuffer.TYPE_INT and no alpha.

A ComponentColorModel can be used in conjunction with a ComponentSampleModel, a BandedSampleModel, or a PixelInterleavedSampleModel to construct a BufferedImage.