Image analysis – Image transformation or preprocessing – Changing the image coordinates
Reexamination Certificate
2000-11-06
2004-01-27
Ahmed, Samir (Department: 2623)
Image analysis
Image transformation or preprocessing
Changing the image coordinates
C382S232000
Reexamination Certificate
active
06683997
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to an apparatus and a method of transforming a compressed image with one resolution to a compressed image with a different resolution at an orthogonal transformation coefficient level.
In recent years, a large number of image compression techniques using DCT (Discrete Cosine transform) have been proposed. For example, H.261 moving image coding for video conferences of CCITT (Comite Consultatif Internationable Telegraphique Expert Group), MPEG (Moving Picture Experts Group) moving image coding of ISO (International Standards Organization), JPEG (Joint Photographic Expert Group) still image coding of ISO and CCITT, and the like can be named.
Generally, to transform the resolution of a discrete-cosinetransformed image, the image is restored to the non-compression state using inverse DCT (IDCT) and thinning-out processing, etc., is performed for the image for transforming the resolution of the image, then DCT is again executed.
However, the resolution transformation method requires the image processing of DCT and inverse DCT in addition to the resolution transformation processing, and, therefore, high-speed processing cannot be performed. Applications of transforming the resolution of a discret ecosine transformed image at high speed, such as a program picture thumbnail function of scaling down a plurality of program pictures received in a digital broadcast (MPEG) and displaying a plurality of programs on one screen, are likely to be necessary in the future.
The method of transforming the resolution of an image at high speed is disclosed, for example, in JP-A-8-180194. In the method, as shown in
FIG. 1
in JP-A-8-180194, image coded data is decoded into orthogonal transformation coefficient data by variable-length decoding means and inverse quantization means, the resolution of the data is transformed at an orthogonal transformation coefficient level by transformation coefficient combining means for generating one transformation coefficient block from two or more (n) transformation coefficient blocks, and the orthogonal transformation coefficient data is again coded by quantization means and variable-length coding means, whereby the effect of being capable of transforming the resolution at high speed at the orthogonal transformation coefficient level can be produced without involving image processing of orthogonal transformation and orthogonal inverse transformation.
As the method of transforming the resolution at the orthogonal transformation coefficient level, the case where one orthogonal transformation block is calculated from two×two orthogonal transformation blocks (reduction to a half in a longitudinal direction and a half in a lateral direction) as shown in
FIG. 25
is taken as an example.
In the description to follow, two-dimensional DCT (IDCT) with eight rows and eight columns is used as orthogonal (inverse) transformation and a block consisting of A rows and B columns of DCT coefficients is defined as an A×B DCT block, anon-compression block with A rows and B columns is defined as an A×B non-compression block, and two-dimensional DCT (IDCT) with A rows and B columns is defined as DCT A×B (IDCT A×B). A transformation matrix for generating one 8×8 DCT block from two×two of 8×8 DCT blocks is calculated by the following method:
1. Two×two of 8×8 DCT blocks are subjected to IDCT 8×8 to restore to 16×16 non-compression blocks.
2. The 16×16 non-compression blocks are subjected to DCT 16×16 to generate 16×16 DCT blocks.
3. Low-frequency 8×8 DCT areas of the 16×16 DCT blocks are adopted as 8×8 DCT blocks after transformation.
4. A transformation matrix for directly performing computations 1 to 3 is calculated.
Generally, DCT (IDCT) computation can be represented as matrix computation and thus the above-mentioned transformation matrix can also be calculated. Therefore, resolution transformation at a DCT coefficient level can be accomplished.
In the related art, however, resolution transformation processing is performed using all orthogonal transformation coefficients before resolution transformation, thus a problem of an increase in the computation amount for performing resolution transformation in response to the longitudinal and lateral change ratios occurs.
The transformation matrix or transformation expression of resolution transformation must be generated in response to the longitudinal and lateral change ratios; this is a problem.
A method of executing resolution transformation of one-dimensional orthogonal transformation blocks in response to the longitudinal and lateral change ratios of an image is not designed.
To perform matrix computation of resolution transformation using a transformation matrix, a speeding-upmethod using the characteristics of orthogonal transformation is shown concerning the even-numbered'th coefficients of the found orthogonal transformation block, but not shown concerning the odd-numbered'th coefficients.
To generate orthogonal transformation blocks with a plurality of resolutions at the same time, generally transformation matrixes are generated in response to the longitudinal and lateral change ratios and resolution transformation processing is performed separately.
Generally, image compression and decompression processing inverse quantization and quantization processing, but hitherto, a resolution transformation method containing quantization has not been designed.
The transformation expression of resolution transformation containing quantization must be generated in response to the longitudinal and lateral change ratios; this is a problem.
To generate orthogonal transformation blocks with a plurality of resolutions at the same time, hitherto a resolution transformation method containing quantization has not been designed.
SUMMARY OF THE INVENTION
To solve the problems, in the invention, first only low-frequency areas of orthogonal transformation blocks required for resolution transformation are extracted and resolution transformation processing is performed. Thus, the effect of suppressing an increase in the computation amount for resolution transformation in response to the longitudinal and lateral change ratios is produced.
Second, the transformation matrixes or transformation expressions of resolution transformation corresponding to all longitudinal and lateral change ratios are previously stored in nonvolatile memory. Thus, the effect of being capable of decreasing processing of generating each transformation matrix or transformation expression of resolution transformation in response to the longitudinal and lateral change ratios is produced.
Third, if an image made up of one-dimensional orthogonal transformation blocks in a lateral direction, resolution transformation of averaging processing or thinning-out processing among orthogonal transformation blocks is performed in a longitudinal direction, then resolution transformation in orthogonal transformation block in the lateral direction is executed. Thus, the effect of being capable of executing resolution transformation of one-dimensional orthogonal transformation blocks in two-dimensional directions of longitudinal and lateral directions is produced.
Fourth, in one-dimensional orthogonal transformation, the even-numbered'th and odd-numbered'th coefficients of the orthogonal transformation block after resolution transformation are calculated using the nature of orthogonal transformation. In two-dimensional orthogonal transformation, the coefficients on (even-numbered rows, even-numbered columns), (even-numbered rows, odd-numbered columns), (odd-numbered rows, even-numbered columns), and (odd-numbered rows, odd-numbered columns) of the orthogonal transformation block after resolution transformation are calculated using the nature of orthogonal transformation. Thus, the effect of improving the computation processing of resolution transformation is produced.
Fifthly, to generate orthogonal transformation blocks w
Ando Atsushi
Iwasaki Osamu
Komiya Daisaku
Uenoyama Tsutomu
Yamada Kazunori
Ahmed Samir
Bali Vikkram
Matsushita Electric - Industrial Co., Ltd.
Pearne & Gordon LLP
LandOfFree
Apparatus and method for resolution transformation of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Apparatus and method for resolution transformation of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Apparatus and method for resolution transformation of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3266364