Image analysis – Image compression or coding – Shape – icon – or feature-based compression
Reexamination Certificate
2000-12-29
2003-12-09
Chen, WenPeng (Department: 2624)
Image analysis
Image compression or coding
Shape, icon, or feature-based compression
C382S236000, C375S240080, C375S240120
Reexamination Certificate
active
06661928
ABSTRACT:
FIELD OF TECHNOLOGY
This invention relates to an image encoding apparatus, an image decoding apparatus, an image encoding method, an image decoding method, an image encoding program recording medium and an image decoding program recording medium.
PRIOR ART
The image encoding technology has a long history. There has been established excellent standard proposals such as ITU-T H.261, ITU-T H263, ISO MPEG1/2 and so on. Roughly speaking, the image encoding method has two approaches: an encoding method using the orthogonal transform and a prediction encoding method encoding the error of predicted values with the use of the prediction function. Although the encoding method using the orthogonal transform needs complicated calculation, when encoded signals of small bit numbers are obtained, it is possible to keep better picture quality than the prediction encoding method. The ordinary encoding method using orthogonal transform such as JPEG, MPEG and the like utilizes the DCT (Discrete Cosine Transform). Though it is known that DCT enables encoding by a small number of bits, it has own problems in that it needs high-precision multiplication, resulting in complicated calculation, and in that the reversible encoding is impossible. Accordingly, DCT calculation can not be used in the fields in which the reversibility is required.
As opposed to this, the prediction encoding method needs simple calculations and can do the reversible encoding. MMR (Modified Modified Read) used in facsimiles is famous as an image coding method having reversibility. MMR is used according to CCITT Rec.T6 “Facsimile Coding Schemes and Coding Control Functions for Group 4 Facsimile Apparatus”. In this method, the difference value in the horizontal direction between the change points of pixel values on the immediately previous already-coded scanning line and the change points of pixel value on the not-yet-coded scanning line is variable-length encoded. MMMR (Modified MMR) which is a further improved MMR is used as the evaluation model for MPEG4 (ISO/IEC/JTC/SC29/WG11 N1277, July 1996).
Incidentally, if image signals are separated into the objects and then the objects are processed as arbitrary shapes, image can be operated and synthesized, object by object, which leads to the effective signal-transmission. For applications which restrict bit number, by using such information, it is possible to selectively assign priority to important objects to transmit and record the same. However, the prior art technology has not taken into account the encoding of objects having arbitrary shapes. And the standardization of coding for image signals having arbitrary shapes has been proceeding in the ISO MPEG4. In MPEG4, the evaluation model called VM3.0 (printed in ISO/IEC/JTC1/SC29/WG11 N1277) is created, which is now a unique image encoding method that can encode image signals having arbitrary shapes.
An image signal having arbitrary shapes ordinarily consists of the shape information indicating the shape of an object and the pixel value information (color information) representing pixel values within an object. Concerning the shape information, the two-valued shape information indicating whether each pixel is significant(on the inside of the shape) or insignificant(on the outside of the shape), or the transparency information indicating the ratio (how much the object occludes the background) of respective pixels which is used in synthesizing with other images. When the transparency has only two levels, 0% and 100%, the shape information is identical to the transparency information and thereby the arbitrary-shape-having image signal is represented by the two of the two-valued shape information and the pixel value information.
FIG. 53
is a drawing for explaining these information. The transparency information is an information representing how much ratio of each pixel is used for synthesis when a fish shown in FIG.
53
(
a
) is synthesized with the other image. In FIG.
53
(
b
), there is shown the value of transparency information in the horizontal scanning line indicated by a dotted line in the figure. The outside of the fish is perfectly transparent. Here, the transparency 0 is defined as being perfectly transparent for simplification. Hence, on the outside of the fish the transparency information has a value of 0, while on the inside of the fish it has a value of non-0.
FIG.
53
(
c
) shows the transparency which is made two-valued as having two of 0 and non-0. In FIG.
53
(
c
), the pixels having the non-0 transparency require encoding of the pixel value information, while the pixels having the 0 transparency do not need the pixel value information, so that the two-valued transparency information is very important to the pixel value information encoding. On the other hand, the component of the transparency information which cannot be represented by two-valued information, as shown in FIG.
53
(
d
), is multi-valued information which is called gray scale. The shape information represented by multi-valued information as described above can be treated by the waveform encoding similar as that for the pixel value information.
While performing the image encoding, the intra-frame encoding based on the spacial correlation or the temporal correlation is separately used, both of the two are employed. In the inter-frame encoding, the motion in the close frames is detected, and the motion compensation is carried out for the detected motion. The motion vector is generally used for the motion compensation. In the above-mentioned VM3.0, the intra-frame encoding and the inter-frame encoding are adaptively switched each other block by block, and the motion compensation similar as in MPEG1/2 is carried out, whereby the efficiency of encoding is improved.
As described above, when performing encoding to the image consisting of the shape information and the pixel value information, if the motion compensation encoding of a shape information is carried out using the motion vector of the pixel value information for the shape information to be used for the image synthesis, the efficiency of encoding is further improved than when the shape information is directly encoded. This is reported by ISO/IEC/JTC1/SC29/WG11 N1260 March 1996.
Further, when the motion detection and motion compensation are executed, it is considered that it is efficient that the shape information is separated into the two-valued shape information component and the multi-valued information component, and the multi-valued information component as well as the pixel value information are subjected to the same waveform encoding together, which has been actually practiced.
In the above-described prior art image encoding and the image decoding accompanying to this, there exist the following problems.
Though MMR encoding is a representative one of the reversible (loss-less) encoding as described above, because of the reversibility, it is impossible to largely improve the compression rate by allowing the visually less-important picture-quality degradation.
In addition, MMR is an intra-frame encoding method, and does not take into account the improvement of the compression rate by utilizing the inter-frame correlation. In MMR and MMMR which a modified version of MMR, only the difference between the change point of the current scanning line and the change point of the immediately previous scanning line is utilized, and the redundancy by the correlation as a straight line in the vertical direction is not sufficiently removed. Accordingly, the encoding efficiency is good when the change of the pixel value happens along the scanning line, but the encoding efficiency is bad when the change of the pixel value does not happen along the scanning line. MMR and MMMR also includes the horizontal encoding mode which does not utilize the correlation in the vertical direction at all in order to encode the pixels which can not be encoded as the difference of the change point of the immediately previous scanning line. This horizontal encoding mode has a room for further improving the efficiency with the u
Chen WenPeng
Parkhurst & Wendel L.L.P.
LandOfFree
Image encoding apparatus, image decoding apparatus, image... 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 encoding apparatus, image decoding apparatus, image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image encoding apparatus, image decoding apparatus, image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3114502