The present invention relates to the field of image processing. More particularly, the present invention relates to blending two sources of image data to form a resulting image.
A digital image is made up of many pixels that are represented in a computer system as digital values. The pixels are typically arranged in columns and rows such that the collected pixels form images to be perceived by a user. Pixels can be displayed as color or gray-scale images (also called black and white images). In a gray-scale image, each pixels is represented by a luminance (also called intensity) value. For example, where luminance is expressed using eight digital bits, each pixel is represented by a single unsigned byte with a range of 0 to 255, where 0 specifies the darkest pixel, 255 specifies the brightest pixels, and intermediate values specify intermediate luminance. Moreover, images can be represented in color format wherein chrominances and luminances are specified. Chrominance (or color) can be represented using a collection of colors that when mixed together generate a perceived color. For example, the colors red (R), green (G), and blue (B) can be used to display many different colors. Depending on the chromatic content of the individual colors, a color gamut is available from which colors can be generated.
Digital image technology finds many applications including television, video and computer graphics. In these types of applications, it can be straightforward to display any given image; however, many possibilities exist when it becomes necessary to display two or more images over the same display area. Whereas to display either of two images as if it were placed on top of the other can be a simple task, it can, however, be the case that a mix or blending of the images is necessary. In the field of art to which it pertains, this mixing or blending is often achieved through a technique called alpha blending. Through the use of alpha blending, two or more images can be individually displayed, but can also be partially displayed with one image dominating the other.
In computer graphics applications, for example, in computer games, blending of images is very important, however, for complex or fast-paced graphics, alpha blending calculations can become computationally intensive. In certain applications, a screen can contain millions of individual pixels where each pixel must be refreshed every 1/60th of a second. Moreover, many or all such pixels may require alpha blending calculation. Thus, there is a need for a method for fast alpha blending calculations in systems incorporating complex digital graphics. There is a further need for an apparatus that can produce fast alpha blending calculations through reduced hardware or reduced complexity hardware, software, or firmware.
The present invention discloses an apparatus and method for performing alpha blending calculations in a fast and efficient manner. Moreover, when implemented as an integrated circuit, an apparatus according to the present invention occupies reduced chip area. In an embodiment of the invention, an apparatus is disclosed for performing alpha blending calculations on a first digital image value, X, and a second digital image value, Y, according to a digital value, alpha (α), that approximates the quantity, αX+(1−α)Y. The α value has a fractional part of n digits, a most significant bit of α1, and a least significant bit of αn−1, wherein the order of αi is related to i. The apparatus comprises a plurality of multiplexers and an adder. Each of the plurality of multiplexers is configured to receive a corresponding bit value, αi, of the digital value α. Moreover, each of the plurality of multiplexers is configured to receive the first and second digital image values. Still further, each of the plurality of multiplexers is configured to direct to outputs of the plurality of multiplexers either the first or second digital image value responsive to the bit value, α1, of the digital value α. Each of the outputs of the plurality of multiplexers is shifted left according to the order of the bit values, αi, of α to produce a plurality of left-shifted outputs. The adder is configured to receive the plurality of left-shifted outputs and is further configured to produce the sum of the left-shifted outputs.
One embodiment of the present invention is an apparatus for performing alpha blending calculations on a first digital image value and a second digital image value according to a digital alpha (α) value having a fractional part of n digits and having a most significant bit of α1 and a least significant bit of αn−1, wherein the order of αi is related to an ith digit of α. The apparatus includes a plurality of multiplexers, a plurality of shifters, and an adder. Each of the plurality of multiplexers is configured to receive a bit value, αi of the digital value α, to receive the first and second digital image values, and to direct to outputs of the plurality of multiplexers either the first or second digital image value responsive to the bit value, αi, of the digital value α. Each shifter has a set of inputs connected to the outputs of one of the multiplexers to receive the selected digital image value, where each shifter left-shifts the selected digital image value according to the multiplexer to which the shifter is connected. The adder is configured to receive the plurality of left-shifted outputs from the shifter and further configured to produce the sum of the left-shifted outputs from all of the shifter.
Another embodiment of the present invention is a method for performing alpha blending calculations on a first digital image value X and a second digital image value Y according to a digital alpha (α) value having a fractional part of n digits and having a most significant bit of α1 and a least significant bit of αn−1 and wherein the order of αi is i. The method includes (a) receiving the first and second digital image values, (b) receiving the α value, (c) selecting the first digital value responsive to a logical high of the bit value, αi, of a, or otherwise selecting the second digital value responsive to a logical low of the bit value, αi, of α, where the selected first or second digital value is a selected digital value, (d) shifting the selected digital value in a shifter left according to the order of the bit values, αi, of α to generate a plurality of left shifted digital values, and (d) adding the plurality of left-shifted digital values from the shifters to generate a result, R.
In another embodiment of the invention, the first digital image value is represented using a first number of bits and the second digital image value is represented using a second number of bits. Also, the sum of the left-shifted outputs can be represented using a third number of bits. In an embodiment, the third number of bits is greater than or equal to the sum of the first number and the second number plus one. In another embodiment, the first and second left-shifted digital image values are produced by computer controlled bit operations. Also, the adding step can be performed using a hardware-implemented adder.
Many other embodiments or variations are possible as will be appreciated upon an understanding of the present disclosure.
The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention:
In order to fully understand the present invention, it is important to understand the concept of alpha blending. Shown in
Shown in
In graphics applications, a resulting pixel value, R (e.g., a pixel from within area 108), is produced from a blend of a first pixel value, X (e.g., a pixel from image 104), and a second pixel value, Y (e.g., a pixel form image 106), through the use of the following equation:
R=αX+(1−α)Y
where α is a number ranging from 0 to 1. From this linear equation, it can be seen that where α=1, the resulting pixel value is the same as the first pixel value (i.e., R=X, when α=1); it is also seen that where α=0, the resulting pixel value is the same as the second pixel value (i.e., R=Y, when α=0). It can also be seen that where α=0.5, the resulting pixel value is a blend of both images (i.e., R=0.5X+0.5Y). This is the condition shown for area 108 of
Many graphic systems are implemented on digital computers where information including graphical information is communicated in binary format. Thus, information is represented as collections of ones and zeros. Moreover, mathematical calculations are performed using a binary counting system. In order to fully appreciate the present invention, it is important to understand how the prior art performed alpha blending calculations.
Shown in
Shown in
Whereas prior art implementations such as those shown in
Given a positive number α in base r with an integer part of m digits and a fraction part of n digits, the (r−1)'s complement of α is defined as
{overscore (α)}=rm−r−n−α.
We can apply this to the particular case where r=2 (i.e., a binary or digital system), an integer part of zero digits and a fraction part of n digits. Thus, the 1's complement of α is defined as
{overscore (α)}=1−r−n−α.
Rewriting this result, we have
{overscore (α)}=(1−α)−r−n.
For n of sufficient size, it is observed that the term r−n becomes very small such that the 1's complement of α is approximated as
{overscore (α)}≈(1−α).
Whereas the calculation of the 1's complement of α can be mathematically complex, in actual implementation, the 1's complement is achieved by simply reversing each bit of a digital number. For example, for the digital number α=0.01101001 (i.e., α=0.41015625 in base 10), the 1's complement is achieved by simply reversing each bit resulting in the 1's complement {overscore (α)}=0.10010110 (i.e., {overscore (α)}=0.5859375 in base 10; note that for the purposes of the present disclosure, leading zeros will be omitted when referring to digital numbers if no bit is present to represent the leading zero). We see that {overscore (α)}=0.5859375 closely approximates the true value of 1−α=0.58984375 where the difference is equal to 2−8=0.00390625. Thus, because the complement of α can be easily obtained, so can an approximation of the quantity 1−α. A fast and efficient manner for calculating the quantity (1−α) is developed from this approximation. Moreover, a fast and efficient manner for performing alpha blending is developed.
Recall that the alpha blending calculation takes two quantities X and Y and blends them together according to a value α to produce a mixed value R according to the equation:
R=αX+(1−α)Y.
From the approximation for {overscore (α)}, we obtain an estimate for the value R:
R≈αX+{overscore (α)}Y
We can therefore rewrite the equation for R as
where
α1 (sometimes also written as α(1)) is the most significant bit (MSB) of α,
αn−1 (sometimes also written as α(n−1) is the least significant bit (LSB) of α,
{overscore (α)}1 (sometimes also written as {overscore (α)}(1)) is the MSB of {overscore (α)},
{overscore (α)}n−1 (sometimes also written as {overscore (α)}(n−1) is the LSB of {overscore (α)}; and
Ri is a partial sum corresponding to αi or {overscore (α)}i.
From the approximation for R above, the number i corresponds in a known way to the order of the bit value αi. Consider, for example, the partial sum
Ri=αi2−i−1X+{overscore (α1)},2−1−1Y
where i=2, α2=1 and {overscore (α)}2=0 such that we have the following partial sum
Now consider, for example, the condition where i=3, α3=0, and {overscore (α)}3=1 such that we have the following partial sum
It is, therefore, observed that the result R is obtained as the sum of left-shifted digital quantities of the quantities X and Y, where the quantities X and Y are left-shifted according to the order of the corresponding bit of α or {overscore (α)}, respectively. Because the quantities α and {overscore (α)} are complements of each other, it is further observed that any partial sum, Ri, is obtained from either a shifted quantity of X or Y, but not both.
An example will assist in understanding the concepts just described. Assume that X and Y are 8-bit numbers. For the purposes of clarity the subscript 8 (i.e., X8 and Y8) will be used to remind the reader that these are 8-bit values; moreover, the subscript 8 will be used with the value zero (i.e., 08) to denote an 8-bit zero value. Assume that α is an eight-bit number less than 1 and greater than or equal to zero such that {overscore (α)} is also less than 1 and greater than or equal to zero. In proceeding, specific values of α and {overscore (α)} will be used to illustrate key features of the present invention, but one of skill in the art will understand its broad implications. If α=0.011001, then {overscore (α)}=0.10010110. The product α·X8 is, therefore, conceptually obtained as follows:
Note that in this result, 08,L7+X8,L6+X8,L5+08,L4+X8,L3+08,L2+08,L1+X8,L0, the subscript Lx has been included to denote the number of left shifts associated with a value (e.g., L7 indicates a left shift of 7 bits). Thus, to obtain the product α·X8 requires shifting of the value X8 according to the digital representation of α. Where a 1 is present in the digital value αi, the digital value X8 is shifted left according to the order of αi such that the final product α·X8 is the sum of all the shifted values; zero values do not add to the complexity of the present calculation, but are included here as place holders toward obtaining a final solution. Placement of the decimal point must account for the fractional number α and the fractional part of the value X8 as is known to those of skill in the art. Similarly, with the 1's complement of α, the product {overscore (α)}·Y8 is obtained as follows
The quantity R can, therefore, be obtained by adding these calculated values:
Observe that the values of X8,Ly and Y8,Ly (where y is a value between 0 and 7) are never added to each other such that only one or the other of the shifted values, X8,Ly and Y8,Ly, can be input to an adder to calculate the resulting sum. The adder used in such a function would need to account for possible carries of the resulting sum, Y8,L7+X8,L6+X8,L5+Y8,L4+X8,L3+Y8,L2+Y8,L1+X8,L0.
It is further observed that an approximate value for R can be obtained by selecting either X or Y according to the digital representation of α, shifting the selected value according to the order of αi and then adding the selected and shifted quantities (note that here no subscripts are used on X and Y because the more general case is being discussed).
Shown in
The method 300 of
Recall that various left shifts of X and Y were required to obtain the value R=αX+{overscore (α)}Y. Left shift units 412-0 through 412-7 provide this shifting function. In actual implementation, left shift unit 412-0 can be omitted, as no shift is required. It is nonetheless included so as to highlight the symmetry of implementation. In an embodiment of the invention, left shift units 412-0 through 412-7 are shift registers. It is further noted that the function of left shift units 412-0 through 412-7 can be readily implemented in various ways without introducing a separate device. For example, shifting a value left can be achieved through wiring a device with shifted inputs or outputs. Moreover, where a digital computer is in use, the digital computer can readily provide shifted values of either X or Y.
The outputs of left shift units 412-0 through 412-7 are then directed to adder 414. One of skill in the art will understand that adder 414 can be implemented in many forms. In an embodiment of the invention, adder 414 is implemented as a carry-save adder (CSA). The output at 418 of adder 414 is therefore the desired result, R. In the embodiment being described, adder 414 provides 17 bit lines to represent the result R.
To further highlight features of the present invention,
Another embodiment of the present invention is shown in
It is noted that implementations of the present invention shown in
The foregoing descriptions of specific embodiments of the present invention have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed. Many modifications and variations are possible in light of the above teachings without deviation from the scope of the claims set out below. The embodiments were chosen and described in order to best explain the principles of the invention and its practical application, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated.
Number | Name | Date | Kind |
---|---|---|---|
5517437 | Yamashita et al. | May 1996 | A |
5754185 | Hsiao et al. | May 1998 | A |
6466220 | Cesana et al. | Oct 2002 | B1 |
6567096 | Oka et al. | May 2003 | B1 |
6727905 | Narita | Apr 2004 | B1 |
Number | Date | Country | |
---|---|---|---|
20040005099 A1 | Jan 2004 | US |