WO2006044880A1

WO2006044880A1 - Methods for determining block averages for film grain simulation

Info

Publication number: WO2006044880A1
Application number: PCT/US2005/037405
Authority: WO
Inventors: Jeffrey Allen Cooper; Joan Llach; Cristina Gomila
Original assignee: Thomson Licensing
Priority date: 2004-10-18
Filing date: 2005-10-17
Publication date: 2006-04-27
Also published as: MX2007004365A; US9953401B2; JP4764883B2; JP2008517531A; EP1803091A1; EP1803095B1; RU2372658C2; US20150187387A1; US8014558B2; MY151714A; JP2008517538A; JP5214246B2; HK1104364A1; BRPI0516471A; CN101040300B; WO2006044684A1; CN101040300A; ZA200703560B; EP1803091B1; KR20070083688A

Abstract

The present invention provides methods for determining block averages in film grain simulation including determining block averages during a display process and determining block averages during a decoding process. The methods of the present invention exhibit different characteristics in terms of memory requirements and computational cost. More specifically, the first method uses no external memory, but requires either extra reads of the blocks, or internal memory in the display pipeline, while the second method requires extra memory bandwidth and extra external memory (e.g., RAM).

Description

METHODS FOR DETERMINING BLOCK AVERAGES FOR FILM GRAIN SIMULATION

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims priority to United States Provisional Patent Application Serial No. 60/619,632, filed October 18, 2004, which is herein incorporated by reference in its entirety.

FIELD OF THE INVENTION

The present invention generally relates to film grain simulation and, more particularly, to methods for determining block averages in film grain simulation.

BACKGROUND OF THE INVENTION

Film grain forms in motion picture images during the process of development. Film grain is clearly noticeable in HD images and becomes a distinctive cinema trait that is becoming more desirable to preserve through the whole image processing and delivery chain. Nevertheless, film grain preservation is a challenge for current encoders since compression gains related to temporal prediction cannot be exploited. Because of the random nature of the grain, visually lossless encoding is only achieved at very high bit-rates. Lossy encoders tend to suppress the film grain when filtering the high frequencies typically associated with noise and fine textures.

In the recently created H.264 I MPEG-4 AVC video compression standard, and in particular in its Fidelity Range Extensions (FRExt) Amendment 1 (JVT- K051 , ITU-T Recommendation H.264 I ISO/IEC 14496-10 International Standard with Amendment 1 , Redmond, USA, June 2004), a film grain Supplemental Enhancement Information (SEI) message has been defined. Such a message describes the film grain characteristics regarding attributes like size and intensity, and allows a video decoder to simulate the film grain look onto a decoded picture. The H.264 I MPEG-4 AVC standard specifies which parameters are present in the film grain SEI message, how to interpret them and the syntax to be used to encode the SEI message in binary format. The standard does not specify, however, the exact procedure to simulate film grain upon reception of the film grain SEI message. Film grain simulation is a relatively new technology used in post-production to simulate film grain on computer-generated material, as well as during restoration of old film stocks. For this kind of applications, there exists commercial software in the market like Cineon®, from Eastman Kodak Co, Rochester, NY, and Grain Surgery™, from Visual Infinity. These tools require user interaction and are complex to implement, which makes them unsuitable for real-time video coding applications. Furthermore, none of these tools has the capability to interpret a film grain SEi message as specified by the H.264 / AVC video coding standard or the like.

SUMMARY OF THE INVENTION

The present invention advantageously provides methods for determining block averages in film grain simulation.

In one embodiment of the present invention, a method for determining block averages in film grain simulation includes computing the block averages during the decoding process, providing efficient computation and low memory bandwidth requirements.

In an alternate embodiment of the present invention, a method for determining block averages in film grain simulation includes computing block averages in raster-scan order during the display process. This embodiment allows using typical video decoders and at the same time provides a memory efficient implementation.

BRIEF DESCRIPTION OF THE DRAWINGS The teachings of the present invention can be readily understood by considering the following detailed description in conjunction with the accompanying drawings, in which: FIG. 1 depicts a high level block diagram of a video decoder subsystem having film grain simulation capabilities in accordance with one embodiment of the present invention;

FIG. 2 depicts a high level block diagram of an embodiment of an interface controller suitable for use in the video decoder subsystem of FIG. 1 ;

FIG. 3 depicts a high level block diagram of a typical arrangement of the film grain database of FIG. 1 ;

FIG. 4 depicts a flow diagram of a method for the computation of block averages during the display process in accordance with one embodiment of the present invention; and

FIG. 5 depicts a flow diagram of a method for the computation of block averages during the decode process in accordance with one embodiment of the present invention.

It should be understood that the drawings are for purposes of illustrating the concepts of the invention and are not necessarily the only possible configuration for illustrating the invention. To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the figures.

DETAILED DESCRIPTION OF THE INVENTION

The present invention advantageously provides methods for determining block averages in film grain simulation. Although the present invention will be described primarily within the context of a video decoder subsystem for application in, for example, IC designs for consumer HD DVD players, the specific embodiments of the present invention should not be treated as limiting the scope of the invention. It will be appreciated by those skilled in the art and informed by the teachings of the present invention that the concepts of the present invention can be advantageously applied in any film grain simulation processes in, for example, media player/receiver devices, decoders, set-top boxes, television sets or the like.

FIG. 1 depicts a high level block diagram of a video decoder subsystem having film grain simulation capabilities in accordance with one embodiment of the present invention. The video decoder subsystem 100 of FIG. 1 illustratively comprises a video decoder (illustratively a H.264 decoder) 106, a video display and graphics engine 108, a host interface 110, an interface controller (illustratively a RAM interface controller) 112, and a local memory (illustratively a RAM memory) 114 implemented as a film grain cache for storing a small subset of the film grain patterns in the remote film grain database 104. FIG. 1 further depicts a host CPU 102 and a remote permanent storage program memory comprising a remote film grain database 104. Although in the video decoder subsystem 100 of FIG. 1 , the host CPU 102 and the remote film grain database 104 are depicted as comprising separate components, in alternate embodiments of the present invention, the remote film grain database 104 can be located in a permanent memory of the CPU 102. Furthermore, although in the video decoder subsystem

100 of FIG. 1 , the video decoder 106, the video display and graphics engine 108, the host interface 100, and the interface controller 112 are depicted as comprising separate components, in alternate embodiments of the present invention, the video decoder 106, the video display and graphics engine 108, the host interface

100, and the interface controller 112 can comprise a single component and can be integrated in a single integrated system-on-chip (SoC) design. Furthermore, although in the video decoder subsystem 100 of FIG. 1 , the means for storing the film grain patterns are depicted as a local memory 114 (cache) and a remote film grain database 104, in alternate embodiments of the present invention, substantially any accessible storage means may be implemented to maintain a subset of the film grain patterns and the total number of film grain patterns. Such means may include storage disks, magnetic storage media, optical storage media or substantially any storage means. In addition, one or more storage means may be implemented for each of the storage devices. Even further, although the film grain database 104 of FIG. 1 is depicted as being located remotely from the memory 114, in alternate embodiments of the present invention, the film grain patterns storage means may be located in close proximity or at great distances from each other.

FIG. 2 depicts a high level block diagram of an embodiment of an interface controller suitable for use in the video decoder subsystem 100 of FIG. 1. The interface controller 112 of FIG. 2 comprises a processor 210 as well as a memory 220 for storing control programs, algorithms and the like. The processor 210 cooperates with conventional support circuitry 230 such as power supplies, clock circuits, cache memory and the like as well as circuits that assist in executing the software routines stored in the memory 220. As such, it is contemplated that some of the process steps discussed herein as software processes may be implemented within hardware, for example, as circuitry that cooperates with the processor 210 to perform various steps. The interface controller 112 also contains input-output circuitry 240 that forms an interface between the various respective functional elements communicating with the interface controller 112.

Although the interface controller 112 of FIG. 2 is depicted as a general purpose computer that is programmed to perform various control functions in accordance with the present invention, the invention can be implemented in hardware, for example, as an application specified integrated circuit (ASIC). As such, the process steps described herein are intended to be broadly interpreted as being equivalently performed by software, hardware, or a combination thereof.

Referring back to the subsystem 100 of FIG. 1 , the remote film grain database 104 is typically relatively large. In one embodiment of the present invention, the H.264 video decoder 106, the video display and graphics engine

108, the host interface 110, the interface controller 112, and the local memory 114 comprise components of an HD DVD player. Film grain patterns from the remote film grain database 104 are needed to be accessed at the sample rate of, for example, the HD DVD player. Therefore, fast access to the large film grain database 104 is necessary. In the subsystem 100 of FIG. 1 in accordance with the present invention, only a small portion of the remote film grain database 104 is used during Supplemental Enhancement Information (SEI) film grain periods, which are leveraged to develop a caching technique to reduce complexity.

More specifically, the film grain simulation process of FIG. 1 requires the decoding of film grain SEI messages, conveyed in the International Standard ITU- T Rec. H.264 I ISO/IEC 14496-10 bit-streams as specified by Amendment 1 (Fidelity Range Extensions), which are both herein included by reference in their entireties. In one embodiment of the present invention, film grain SEI messages are sent preceding I (intra-coded) pictures, and only one film grain SEI message precedes a particular I picture.

In one embodiment of the present invention, the remote film grain database 104 of film grain patterns is composed of 169 patterns of 4,096 film grain samples, each representing a 64x64 film grain image. For example, FIG. 3 depicts a high level block diagram of a typical arrangement of the film grain database of FIG. 1. FIG. 3 depicts a 64x64 sample film grain pattern with i_offset in the x-axis and j_offset in the y-axis. FIG. 3 further depicts the 169 film grain patterns of the various types.

In the film grain database 104, each film grain pattern is synthesized using a different pair of cut frequencies according to a frequency filtering model of the standard specifications. The cut frequencies transmitted in the SEI message are used to access the remote film grain database 104 of film grain patterns during the film grain simulation process. The film grain database 104 is' stored in ROM, Flash, or other permanent storage device, such as the film grain database 104 of the video decoder subsystem 100 of FIG. 1 , and typically does not change. The film grain database 104 contains random film grain patterns in a very large variety of film grain shapes and sizes. However, for a specific video content sequence only a small subset of this database is actually needed to effectively simulate film grain. The specification limits the number of film grain patterns to a small subset for any SEI message period. Therefore, the present invention implements a small film grain cache, such as the local memory 114, which is updated on receipt of SEI messages. Typically, the remote film grain database 104 is stored in the permanent storage of the host CPU 102 or at the site of the host CPU 102. However, it is the video decoder 106 and the video display and graphics engine 108 that need fast access to the film grain database 104. As such, and in accordance with the present invention, the local memory 114 is provided for fast access to at least a subset of the film grain patterns. That is, at least a small subset of the film grain patterns needed or most implemented by the existing SEI message period is transferred to and stored in the local memory 114. In one embodiment of the present invention, the local memory 114 is large enough to store the entire film grain database 104. In such an embodiment, the video decoder 106 and the video display and graphics engine 108 have immediate and fast access, via the controller 112, to all of the available film grain patterns originally stored in the remote film grain database 104. In addition, such an embodiment of the present invention has the advantage that the film grain cache in the local memory 114 does not have to be updated on receipt of an SEI message. Such an embodiment, however, has the disadvantage that more memory (e.g., RAM) is required. In some implementations, however, such large memory space (e.g., RAM) is already available.

In an alternate embodiment of the present invention, the local memory 114 is only large enough to store a subset of the film grain database 104. In such an embodiment, on each receipt of an SEI message, the controller 112 initiates ah examination of the cache of the local memory 114 to determine if any of the subset of film grain patterns already in the local memory 114 needs to be replaced with different film grain patterns in the remote film grain database 104 selected in the new SEI message. An advantage of this technique is a smaller local memory 114 allocation. A disadvantage is that the cache of the local memory 114 must be managed by the controller 112, and in the worst case, a full cache size must be transferred from the remote film grain database 104 to the local memory 114 for each I frame via, for example, the controller 112. In addition, in such an embodiment of the present invention, on device boot up (or reset), the local memory 114 (i.e., the film grain cache) can be pre-initialized by the controller 112 with the most common film grain patterns stored in the remote film grain database 104. That is, the selection of which film grain patterns to store in the local memory 114 depends on empirical data based on what film grain patterns in the film grain database 104 were most often used across a wide selection of film content.

In any event, in the above described embodiments of the present invention, the local memory 114 in accordance with the present invention, in conjunction with the controller 112, enable the video decoder 106 and the video display and graphics engine 108 faster access to the film grain patterns previously only contained in the remote film grain database 104. Referring back to FIG. 1 , in film grain simulation systems, such as the video decoder subsystem 100 of FIG. 1 , the block average (e.g., 8x8 block average) of a decoded picture is required to define the film grain shape, size, and intensity. Therefore, in the video decoder subsystem 100 of FIG. 1 , before accessing the local memory 114 to search for or to select a film grain pattern, the block average needs to be calculated. More specifically, the block average of a decoded picture is a set of values, each one representing the average value of the samples of each one of the blocks that made the picture. For example, if our decoded picture is 32 samples wide and 24 samples high, and the blocks are 8 samples wide and 8 samples high, the block average of the decoded picture can be represented in a matrix form as follows:

Avg1 Avg2 Avg3 Avg4 Block_average = Avg5 Avg6 Avg7 Avg8 Avg9 Avg10 Avg11 Avg12

where Avg1 is the average value of the samples in block 1 , which has its top-left sample at coordinate (0,0) and its bottom-right sample at coordinate (7,7), Avg2 is the average of the samples in block 2, which has its top-left sample at coordinates (8,0), and its bottom-right sample at coordinates (15,7), etc.

In addition, the block average can be computed for each color component separately. For example, if the decoded picture is a color image in the YUV color space (Y with luminance information, U and V with chrominance information), it is possible to compute three block averages for the picture, one per color component. If the sample at coordinates (x,y) for color component c are denoted by decoded_picture[c][x][y], the block average of a block N samples wide and M samples high which top-left sample is at coordinate (Ox, Oy) could be computed as follows:

average = 0 for (i = Ox, i < Ox + N; i++) for (j = Oy, j < Oy + M; j++) average += decoded_picture[c][i][j] endfor endfor average /= N * M.

The previous pseudo-code essentially initializes the block average to 0, then adds to it the value of all the samples in the block and finally divides the resulting value by the number of samples, to obtain the average. This process should be repeated for all three color components. In the case of film grain simulation, where the film grain parameters change on a block basis, the block average of each block (i.e., the average value of the samples of the block) is used to select the appropriate film grain parameters for the block. In accordance with the present invention, the inventors propose herein two options for determining the block average of a decoded picture. In a first embodiment of the present invention, the block average of a decoded picture is determined during the display process. In some implementations, a decoder may already read decoded picture blocks (e.g., 8x8 blocks) from external memory (e.g., RAM) during the display process. In this case and in accordance with the present invention, the block average (e.g., 8x8 block average) is performed in the display pipeline. In accordance with the present invention, it is possible to compute the block average (e.g., 8x8 block average) during the display process even if pixels are read in raster-scan order (i.e. left to right, top to bottom). For example, FIG. 4 depicts a flow diagram of a method for the computation of block averages during the decoding process while reading pixels in raster-scan order in accordance with one embodiment of the present invention.

In the method 400 of FIG. 4, an image is processed in groups of 8 lines, and the block average for all 8x8 blocks in the group of 8 lines is computed in parallel. The method 400 begins by setting the block averages to 0 for the first line in the group. The line is then processed in groups of 8 samples. The first 8 samples are added to the first block average (which corresponds to the first 8x8 block in the group of 8 lines), the second 8 samples are added to the second block, average (which corresponds to the second 8x8 block) and so on. The seven remaining lines are processed in the same way by reusing the same block averages. That is, the first 8 samples in lines 2 to 7 are added to the first block average, the second 8 samples are added to the second block average, and so on. At the end of the seventh line, each block average contains the summation of all the samples of the block. In order to obtain the average value, the summation is divided by the number of samples in the block (64 in this case). Once the block average for all 8x8 blocks of the group of 8 lines is known, film grain can be added to them and the lines are ready to be sent to the display. By decoupling the computation of the block average from the decoding process, this approach allows the use of any decoder (since no changes are required to it). At the same time, by reading pixels in raster-scan order during the display process, efficient memory access techniques can be used, leading to a reduced overhead.

In a second embodiment of the present invention, the block average of a decoded picture is determined during a video decode process (i.e., during the H.264 decode process). During the video compression decode operation the 8x8 block samples are available. For example, FIG. 5 depicts a flow diagram of a method for the computation of block averages during the decode process in accordance with one embodiment of the present invention. The method 500 of FIG. 5 illustrates the combining the decoding of a video frame with the computation of the block average of each block (e.g., 8x8 block). The method 500 of FIG. 5 begins by decoding a macro-block of samples (i.e., most video coding standards, such as MPEG-2 and H.264 | MPEG-4 AVC, use 16x16 macroblocks). Next, the block average of each sub-block (e.g., 8x8 sub-block) is computed and stored into memory. This approach minimizes the amount of data that has to be read from memory because the values required to compute the block average are readily available during the decoding process, reducing the requirements on memory bandwidth.

These block averages are then stored in an external memory (e.g., RAM), such as the local memory 114 of FIG. 1 or a separately provided memory (e.g., RAM), to be accessed during a film grain simulation of the display process. This method of the present invention requires some external storage for each decoded frame before it is displayed. In the worst case scenario, this embodiment requires 146 Kbytes of external memory (e.g., RAM) to store the 8*8 block average values of three decoded frames. In addition, some minimal extra memory bandwidth is required to store and retrieve the block averages (e.g., 8*8 block averages). The two methods for determining block averages in accordance with the embodiments of the present invention described above trade off external memory (e.g., RAM) storage with memory bandwidth. The first method uses no external memory, but requires either extra reads of the 8x8 blocks, or internal memory in the display pipeline. The second method requires a little extra memory bandwidth, and 146 Kbytes of external memory (e.g., RAM). Having described various embodiments for methods, apparatus and systems for film grain simulation (which are intended to be illustrative and not limiting), it is noted that modifications and variations can be made by persons skilled in the art in light of the above teachings. It is therefore to be understood that changes may be made in the particular embodiments of the invention disclosed which are within the scope and spirit of the invention as outlined by the appended claims. While the forgoing is directed to various embodiments of the present invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof. As such, the appropriate scope of the invention is to be determined according to the claims, which follow.

Claims

1. A method for determining block averages in a film grain simulation, comprising: determining a block average in the display pipeline, avoiding external memory usage.

2. The method of claim 1 where the block average is computed for one block at a time.

3. The method of claim 1 where the block average is computed for all the blocks in a row in parallel.

4. The method of claim 1 further comprising: reusing the pixel values used to compute the block average for the film grain simulation process; and, reusing the pixel values used for the film grain simulation process for a display process.

5. A method for determining block averages in a film grain simulation, comprising: determining a block average during the decode process by using external memory.

6. The method of claim 5 further comprising: decoding a set of pixels; compute the block average on the set of decoded pixels.

7. The method of claim 6 wherein the decoded set of pixels is a macro- block.

8. The method of claim 7 wherein the macroblock contains 16x16 samples, and 4 block averages are computed, one per each non-overlapping 8x8 block.