Image analysis – Image compression or coding – Transform coding
Reexamination Certificate
1998-06-30
2002-07-30
Tran, Phuoc (Department: 2621)
Image analysis
Image compression or coding
Transform coding
C382S233000, C382S246000, C382S251000
Reexamination Certificate
active
06427029
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to image compression and reconstruction techniques that can be used for compressing a still picture in the form of image signals obtained by a digital still camera, a digital scanner or the like, transmitting the compressed image signals to a personal computer or the like, storing the compressed image signals, and expanding or decompressing the compressed image signals to restore or reconstruct the image signals on the personal computer or the like.
2. Description of the Prior Art
To capture natural images on personal computers, digital still cameras or digital scanners using single-plate CCD image sensors, have been used. These devices each contain an image compression device and a flash memory to compress and store image signals before transferring them to the personal computer. For the compression, a non-reversible or lossy compression method like JPEG, the international standard defined by the ISO/IEC 10918-1, is normally used.
A typical image compression device utilizing a lossy compression method comprises an interpolating section, an image component correcting section and an image compressing section. The interpolating section interpolates image signals outputted from a CCD image sensor so that each pixel has all the color components, which originally contains only one value of a color component, for example, R (red), G (green) or B (blue).
A typical structure of output signals of the CCD image sensor is shown in
FIG. 8
, while a structure of the image signals after the interpolation is shown in FIG.
9
.
The image component correcting section carries out edge enhancement, white balance adjustment and gamma correction to the interpolated image signals. The edge enhancement is performed to the image signals if necessary. The white balance adjustment is performed to adjust the image signals to the characteristic of the human visual system. The gamma correction is performed to adjust the image signals to the characteristics of a display.
The image compressing section compresses the image signals outputted from the image component correcting section. With JPEG, when the image signals are in the form of the RGB, they are converted into luminance (Y)/color difference (Cb or Cr) signals as shown in FIG.
10
. After the conversion into the luminance/color difference signals, a subsampling process is carried out to decimate the color difference signals. Thereafter, the signals are divided into blocks of 8 pixels by 8 pixels, and DCT (discrete cosine transform) is performed on each block. Then, the transformed coefficients are quantized and zig-zag scanned as shown in FIG.
12
. From this pattern, each set of consecutive zeros is paired with the non-zero number that follows, and Huffman coded to accomplish entropy compression. In
FIG. 12
, each pair of numerical values represents coordinates of a corresponding pixel.
Moreover, JPEG is provided with an “escape code” followed by a pair of fixed length codes in order to avoid bloating of the Huffman table.
The transformed coefficients in each block consist of one DC, coefficient and sixty-three AC coefficients. In the foregoing example, the DC coefficient represents the upper-left coefficient and the AC coefficients represent the remaining sixty-three coefficients.
The DC coefficient is encoded in the following manner:
First, calculate the difference _d_ between the current DC coefficient and the previous DC coefficient. Second, obtain a group number (ssss) and the number of additional bits corresponding to the difference _ d_ from the Huffman table shown in FIG.
11
. The additional bits are used to uniquely identify the DC coefficient, e.g., the group number three has three additional bits. Then, output the Huffman coded group number and the additional bits.
On the other hand, the AC coefficients are first re-ordered, then a sequence of zeros followed by a non-zero number is replaced with a predetermined code, and then additional bits are outputted. For example, if number “7” follows after six zeros, a Huffman code representing “a run of six zeros followed by a value of the group number 3” is used. However, if there are more than 14 consecutive zeros, a code ZRL (zero run length) representing a run of 15 consecutive zeros is used. Further, a code EOB (end of block) is used if the subsequent coefficients are all zeros.
On the other hand, an image reconstruction device in the personal computer or the like receives the compressed image data produced by the image compression device and expands the data by means of reversing the compression routine. Namely, it decodes the two-dimensional Huffman codes, zig-zag positions the coefficients, dequantizes them and carries out the inverse DCT on them to reconstruct the luminance/color difference signals to reproduce the original image.
However, the foregoing conventional image signal processing technique has the following problems:
(1) It requires heavy calculations. Thus, it will take much time for compression and decompression.
As described above, in the conventional technique, the amount of data in the image signals outputted from the CCD image sensor are initially increased through interpolation. Hence, the calculation for the entropy compression will be extensive. For example, when each of the R, G and B signals of 8 bpp (bits per pel) is interpolated and converted to an RGB signal, it will become 24 bpp so that the data amount is tripled before compression. Furthermore, JPEG requires the following three complicated steps:
color space conversion, which uses a product-sum operation of decimal numbers;
DCT, which requires a large amount of calculation; and
quantization of coefficients, which requires time consuming division.
(2) Since the color difference signals are decimated before compression, the image quality will be degraded upon reconstruction.
For example, one format adopted by JPEG is 4:2:2, which decimates the color difference signals and reduces the vertical resolution by ½. Therefore, image signals having 8 bpp originally, and hence 24 bpp (8,8,8) after interpolation, will be degraded to 16 bpp (8,4,4), once the 4:2:2 JPEG format is applied.
(3) Since JPEG doesn't acknowledge the correlation between the adjacent blocks, the efficiency of compression is reduced.
In other words, JPEG encodes and compresses each 8×8 block independently. Generally, the adjacent blocks in the natural image tend to have similar textures so that the coefficients after orthogonal transform, such as DCT, are likely to have similar values. Hence, JPEG is not optimizing the potential efficiency of the compression technique.
(4) Although the two-dimensional Huffman coding used in JPEG is efficient, it can not provide all the combinations of the two elements because it will cause bloating of the Huffman table. That is the reason why the escape code was introduced.
(5) The decoding process is complicated.
JPEG uses the two-dimensional Huffman table with a maximum of 16 bits, or in the case of escape coding, 28 bits. Hence, it requires the use of more than one lookup table to achieve high-speed decoding, which rendering the decoding process complicated.
SUMMARY OF THE INVENTION
Therefore, we present an improved image signal processing method that can eliminate one or more of the foregoing problems.
We further present an improved image signal processing device that can eliminate one or more of the foregoing problems.
We further present a storage medium that allows an electronic device, such as a device including a CPU (central processing unit), to carry out the foregoing improved image signal processing method.
Our improved image signal processing proceeds as follows:
First, we perform an image signal correction on each and every component of a given image, which is composed of the three primary colors—red, blue and green, or the three secondary colors—magenta (blue+red), yellow (red+green), and cyan (green+blue), or any tertiary colors, such as yellow+magenta, magenta+cyan, cyan&pl
Kono Takahiko
Sashida Nobuyuki
Takeuchi Shunichi
Custom Technology Corp.
Lerner David Littenberg Krumholz & Mentlik LLP
Tran Phuoc
LandOfFree
Image signal processing method and device, and storage medium does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image signal processing method and device, and storage medium, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image signal processing method and device, and storage medium will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2835431