Patent Grant
8238613
This invention relates to a technique for simulating film grain in an image.
Motion picture films comprise silver-halide crystals dispersed in an emulsion, which is coated in thin layers on a film base. The exposure and development of these crystals form the photographic image consisting of discrete tiny particles of silver. In color negatives, tiny blobs of dye occur on the sites where the silver crystals form following chemical removal of the silver during development of the film stock. These small specks of dye commonly bear the label ‘grain’ in color film. Grain appears randomly distributed on the resulting image because of the random formation of silver crystals on the original emulsion. Within a uniformly exposed area, some crystals develop after exposure while others do not.
Grain varies in size and shape. The faster the film, the larger the clumps of silver formed and blobs of dye generated, and the more they tend to group together in random patterns. The term “granularity” typically refers to the grain pattern. The naked eye cannot distinguish individual grains, which vary from 0.0002 mm to about 0.002 mm. Instead, the eye resolves groups of grains, referred to as blobs. A viewer identifies these groups of blobs as film grain. As the image resolution becomes larger, the perception of the film grain becomes higher. Film grain becomes clearly noticeable on cinema and High Definition (HD) images, whereas film grain progressively loses importance in Standard Definition (SD) and becomes imperceptible in smaller formats.
Motion picture film typically contains image-dependent noise resulting either from the physical process of exposure and development of the photographic film or from the subsequent editing of the images. Photographic film possesses a characteristic quasi-random pattern, or texture, resulting from physical granularity of the photographic emulsion. Alternatively, simulation of similar pattern can occur in computed-generated images in order to blend them with photographic film. In both cases, this image-dependent noise bears the designation of “film grain.” Quite often, moderate grain texture presents a desirable feature in motion pictures. In some instances, the film grain provides visual cues that facilitate the correct perception of two-dimensional pictures. Film grade often varies within a single film to provide various clues as to time reference, point of view, etc. Many other technical and artistic demands exist for controlling grain texture in the motion picture industry. Therefore, preserving the grainy appearance of images throughout image processing and delivery chain has become a requirement in the motion picture industry.
Several commercially available products have the capability of simulating film grain, often for blending a computer-generated object into natural scene. Cineon® from Eastman Kodak Co, Rochester N.Y., one of the first digital film applications to implement grain simulation, produces very realistic results for many grain types. However, the Cineon® application does not yield good performance for many high speed films because of the noticeable diagonal stripes the application produces for high grain size settings. Further, the Cineon® application fails to simulate grain with adequate fidelity when images become subject to prior processing, for example, such as when the images are copied or digitally processed.
Another commercial product that simulates film grain is Grain Surgery™ from Visual Infinity Inc., which is used as a plug-in of Adobe® After Effects®. The Grain Surgery™ product appears to generate synthetic grain by filtering a set of random numbers. This approach suffers from disadvantage of a high computational complexity.
Thus, a need exists for an efficient film grain simulation technique, which reduces the need for memory bandwidth, and computational effort, thus permitting film grain simulation in cost-sensitive high volume devices, such as set top boxes.
Briefly, in accordance with a preferred embodiment of the present principles, there is provided a method for simulating film grain in an image block of M×N pixels, where N and M are integers greater than zero. The method commences by first computing the average of the pixel values within the block of M×N pixels. A film grain block of M×N pixels is selected from among a pool of previously established blocks containing film grain as a function of the average value of the image block and a random number. Each pixel in the selected film grain block is blended with a corresponding pixel in the image block.
Introduction
The method of the present principles simulates film grain in accordance with film grain information transmitted with an image to which simulated grain is blended. In practice, the transmitted image typically undergoes compression (encoding) prior to transmission via one of a variety of well-known compression schemes, such as the H.264 compression scheme. With the transmitted image compressed using the H.264 compression scheme, transmission of the film grain information typically occurs via a Supplemental Enhancement Information (SEI) message. Pursuant to contributions recently adopted by the standards body responsible for promulgating the H.264 standard, the SEI message can now include various parameters that specify different film grain attributes.
Constraints on the film grain SEI message parameters
The method of the present principles imposes some constraints with regard to the number of parameters and their range of possible values allowed by the H.264 recommendation. TABLE 1 provides a list of such parameters, including a description of their semantics and the constraints imposed by the present principles.
In addition to the previous constraints, the present principles imposes that film grain SEI messages precede I pictures, and only one film grain SEI message can precede a particular I picture. (The presence in the bit stream of slice_type equal to 7 or nal_ref_idc equal to 5, indicates an I picture.)
All the other parameters of the film grain SEI message have no constraint with respect to the standard specification.
Bit-Accurate Implementation of Film Grain Simulation
Film grain simulation in accordance with the present principles occurs in a two-step process. First, generation of a pool of film grain blocks occurs during initialization, as described in greater detail with respect to
The apparatus 10 accomplishes film grain noise initialization using a specified uniform pseudo-random number polynomial generator 12 and using a specified list of 2048 8-bit Gaussian distributed random numbers stored in a look-up table 14 The look-up table 14 stores random numbers in 2's complement form in the range [−63, 63]. The list of Gaussian random numbers appears in the Appendix.
According to the bit-accurate specification of the present principles, generation of the film grain blocks begins with the lowest luminance intensity interval. The uniform random number generator 12 generates an index for the Gaussian random number list stored in the look-up table 14 using a primitive polynomial modulo 2 operator, x18+x5+x2+x1+1. For ease of understanding, the term x(i, s) will indicate the ith symbol of the sequence x, beginning with an initial seed s. The random number seed becomes reset to 1 upon the receipt of each film grain SEI message.
To form an individual 8×8 film grain block, a random block generator 16 reads 8 lines worth of 8 random numbers from the Gaussian random number look-up table 14. A random offset, from the random number generator 12, serves to access each line of 8 random numbers. Each line of the block produced by the block generator 16 is generated as following:
index=x(i, 1)
for n=0 . . . 7, B[i %8][n]=Gaussian_list[(index+n)%2048]
where i increments for each 8×1 block line.
The 8×8 block of random values read by the generator 16 undergoes a transform, typically an integer Discrete Cosine Transform (DCT), performed by an Integer DCT transform block 18. After the DCT transform, the 8×8 random values undergo frequency filtering at a frequency filter 20 in accordance with the cut frequencies specified in the SEI message. Following frequency filtering, the 8×8 random values undergo an inverse DCT transform by an inverse integer DCT block 22. A first scaling block 24 scales the pixels on the top and bottom block lines as follows:
for n=0.7, B′[0][n]=(B[0][n]+1)>>1
for n=0.7, B′[7][n]=(B[7][n]+1)>>1
This process continues until generation of a set of 128 film grain blocks for each luminance intensity interval. Following subsequent scaling by the second scaling block 26, the film grain blocks undergo storage in the film grain pool 28.
Block and Pixel Operations to Simulate Film Grain
A selector block 204 selects a kth film grain block from the pool 28, using the random number generated by the uniform random number generator 16 from the polynomial modulo 128 as the block index. Thus, the noise generator 16, which generates uniformly distributed random numbers using a polynomial for the initialization process described with respect to
previous_index=index
index=x(k, 1)% 128
index ^=(index==previous_index)
Following block selection, a deblocking filter 206 deblocks the pixels on the right most column of the previously selected block and on the left most column of the current block. An adder 208 adds the deblocked film grain block to decoded luma pixels. (Since two horizontally adjacent blocks are required to perform deblocking, there is a 1-block delay between the block selected in 204 and the block added in 208.) A clipper 210 clips the result within the range [0, 255] for display. Note that film grain noise addition only occurs to luma pixels.
Scaling of Cut Frequencies
The parameters in the film grain SEI message of TABLE 1 assume the use of a 16×16 DCT in the simulation process. In particular, horizontal and vertical high cut frequencies, provided by comp_model_value[0][i][1] and comp_model_value[0][i][2], and horizontal and vertical low cut frequencies, provided by comp_model_value[0][i][3] and comp_model_value[0][i][4], serve to filter the transform coefficients of a block of 16×16 values.
In the illustrated embodiment, the use of 8×8 blocks will reduce complexity. Employing an 8×8 block transform using cut frequency parameters based on a 16×16 transform implies that all the cut frequencies require scaling before the grain generation.
The scaling of the cut frequencies occurs as follows:
comp_model_value′[0][i][j]=(comp_model_value [0][i][j]+1)>>1
where j is in the range [1,4]. Note that the scaling constitutes the equivalent of the integer division, rounded up to the nearest integer.
Integer Transform and Variance Scaling
The transform used for the frequency filtering corresponds to an 8×8 integer approximation to the DCT, using the following transformation matrix:
16-bit arithmetic can be used. The forward integer transformation of a given block of random noise is defined as:
{tilde over (B)}=(((T8×B+8)>>4)×T8T+8)>>4
with 11 bits used for {tilde over (B)}.
The inverse integer transform is defined as:
B=(T8T×{tilde over (B)}×T8+128)>>8
with 8 bits used for B.
Following the inverse transform, the block B undergoes scaling as follows, assuming it is in the kth luminance intensity interval,
val=B(i, j)*comp_model_value[0][k][0]
B′(i, j)=(((val−(val>>4)+2log 2
where the operation (val−(val>>4)) compensates the scaling of the integer transform; log 2_scale_factor, transmitted in the SEI message, scales comp_model_value[0][k][0]; and 5 scales the Gaussian numbers provided in the Appendix.
Deblocking Filter 206
As indicated, the film grain simulation apparatus of
application of the deblocking occurs as follows:
b′=(a+(b<<1)+c)>>2
c′=(b+(c<<1)+d)>>2
where b′ and c′ replace the value of the original pixels b and c, respectively. Deblocking of the left and right block edges occurs for every film grain block before addition to the decoded image.
The foregoing describes a technique for simulating film grain in an image.
This application claims the benefit, under 35 U.S.C. §365 of International Application PCT/US2004/033713, filed Oct. 12, 2004, which was published in accordance with PCT Article 21(2) on Apr. 28, 2005 in English and which claims the benefit of U.S. provisional patent application No. 60/511,026, filed Oct. 14, 2003.
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/US2004/033713 | 10/12/2004 | WO | 00 | 10/25/2006 |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2005/039189 | 4/28/2005 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 4663663 | Lee | May 1987 | A |
| 4897775 | Kaassens | Jan 1990 | A |
| 4930023 | Yakame | May 1990 | A |
| 4935816 | Faber | Jun 1990 | A |
| 5028280 | Ihara et al. | Jul 1991 | A |
| 5140414 | Mowry | Aug 1992 | A |
| 5262248 | Ihara et al. | Nov 1993 | A |
| 5335013 | Faber | Aug 1994 | A |
| 5374954 | Mowry | Dec 1994 | A |
| 5406326 | Mowry | Apr 1995 | A |
| 5457491 | Mowry | Oct 1995 | A |
| 5471572 | Buchner et al. | Nov 1995 | A |
| 5475425 | Przyborski et al. | Dec 1995 | A |
| 5550815 | Cloonan et al. | Aug 1996 | A |
| 5629769 | Cookingham et al. | May 1997 | A |
| 5641596 | Gray et al. | Jun 1997 | A |
| 5659382 | Rybczynski | Aug 1997 | A |
| 5687011 | Mowry | Nov 1997 | A |
| 5706361 | Kent et al. | Jan 1998 | A |
| 5709972 | Cookingham et al. | Jan 1998 | A |
| 5767860 | Zimmer et al. | Jun 1998 | A |
| 5805782 | Foran | Sep 1998 | A |
| 5817447 | Yip | Oct 1998 | A |
| 5831673 | Przyborski et al. | Nov 1998 | A |
| 5832172 | Jeon | Nov 1998 | A |
| 5917609 | Breeuwer et al. | Jun 1999 | A |
| 6219838 | Cherichetti et al. | Apr 2001 | B1 |
| 6233647 | Bentz et al. | May 2001 | B1 |
| 6269180 | Sevigny | Jul 2001 | B1 |
| 6327391 | Ohnishi et al. | Dec 2001 | B1 |
| 6373992 | Nagao | Apr 2002 | B1 |
| 6459699 | Kimura et al. | Oct 2002 | B1 |
| 6496221 | Wolf et al. | Dec 2002 | B1 |
| 6559849 | Anderson et al. | May 2003 | B1 |
| 6587509 | Suzuki et al. | Jul 2003 | B1 |
| 6650327 | Airey et al. | Nov 2003 | B1 |
| 6868190 | Morton | Mar 2005 | B1 |
| 6990251 | Edgar | Jan 2006 | B2 |
| 6995793 | Albadawi et al. | Feb 2006 | B1 |
| 7065255 | Chen et al. | Jun 2006 | B2 |
| 7171057 | Wilensky et al. | Jan 2007 | B1 |
| 7738722 | Gomila et al. | Jun 2010 | B2 |
| 7742655 | Gomila et al. | Jun 2010 | B2 |
| 7945106 | Gomila et al. | May 2011 | B2 |
| 7958532 | Paul et al. | Jun 2011 | B2 |
| 20010056568 | Hirotsu et al. | Dec 2001 | A1 |
| 20020034337 | Shekter | Mar 2002 | A1 |
| 20020154140 | Tazaki | Oct 2002 | A1 |
| 20020171649 | Fogg | Nov 2002 | A1 |
| 20030011615 | Tidwell | Jan 2003 | A1 |
| 20030043922 | Kalker et al. | Mar 2003 | A1 |
| 20030206231 | Chen et al. | Nov 2003 | A1 |
| 20030206662 | Avinash et al. | Nov 2003 | A1 |
| 20030218610 | Mech et al. | Nov 2003 | A1 |
| 20040013308 | Jeon et al. | Jan 2004 | A1 |
| 20060072660 | Jia et al. | Apr 2006 | A1 |
| 20060082649 | Gomila et al. | Apr 2006 | A1 |
| 20060083316 | Cooper et al. | Apr 2006 | A1 |
| 20060083426 | Cooper et al. | Apr 2006 | A1 |
| 20060133686 | Gomila et al. | Jun 2006 | A1 |
| 20060182183 | Winger | Aug 2006 | A1 |
| 20060256871 | Boyce et al. | Nov 2006 | A1 |
| 20060291557 | Tourapis | Dec 2006 | A1 |
| 20070002947 | Lu et al. | Jan 2007 | A1 |
| 20070036452 | Llach et al. | Feb 2007 | A1 |
| 20070047658 | Tourapis et al. | Mar 2007 | A1 |
| 20070058866 | Boyce et al. | Mar 2007 | A1 |
| 20070058878 | Gomilla et al. | Mar 2007 | A1 |
| 20070070241 | Boyce et al. | Mar 2007 | A1 |
| 20070104380 | Gomila et al. | May 2007 | A1 |
| 20070117291 | Cooper et al. | May 2007 | A1 |
| 20080252781 | De Waele et al. | Oct 2008 | A1 |
| Number | Date | Country |
|---|---|---|
| 0364285 | Apr 1990 | EP |
| 0373615 | Jun 1990 | EP |
| 0879006 | Dec 1990 | EP |
| 0622000 | Oct 1992 | EP |
| 0766281 | Jan 1997 | EP |
| 0575006 | Oct 1999 | EP |
| 1175091 | Jan 2002 | EP |
| 1215624 | Jun 2002 | EP |
| 1511320 | Mar 2005 | EP |
| 2312124 | Oct 1997 | GB |
| 4-097681 | Mar 1992 | JP |
| 9-073765 | Mar 1996 | JP |
| 8-149466 | Jun 1996 | JP |
| 9-082718 | Mar 1997 | JP |
| 2001357090 | Dec 2001 | JP |
| 2002359846 | Dec 2005 | JP |
| 2007829945 | Oct 2007 | JP |
| WO9314591 | Jul 1993 | WO |
| WO9520292 | Jul 1995 | WO |
| WO 9710575 | Mar 1997 | WO |
| WO 9722204 | Jun 1997 | WO |
| WO 0018109 | Mar 2000 | WO |
| WO-0146992 | Jun 2001 | WO |
| WO-0177871 | Oct 2001 | WO |
| WO 0233958 | Apr 2002 | WO |
| WO 2004095829 | Nov 2004 | WO |
| WO-2004104931 | Dec 2004 | WO |
| WO2004105250 | Dec 2004 | WO |
| WO 2004034518 | Apr 2005 | WO |
| WO 2008039189 | Apr 2005 | WO |
| WO 2006057703 | Jun 2005 | WO |
| WO 2005027046 | Mar 2006 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 20070070241 A1 | Mar 2007 | US |
| Number | Date | Country | |
|---|---|---|---|
| 60511026 | Oct 2003 | US |
