JPH01233892A - Three-dimension picture conversion coding device - Google Patents

Abstract

PURPOSE:To improve a compression rate more than a two-dimension extrapolation prediction discrete sine transform by utilizing correlation relating to elapsed time of a picture data so as to apply compression also in a direction of time base. CONSTITUTION:A memory control means 2 extracts switchingly a block of a prescribed picture element in horizontal and vertical direction from a picture data of each sheet, the two-dimension extrapolation prediction discrete sine transform means 3 applies prediction conversion coding and stores the data subjected to prediction transform coding into a memory 1, the block of prescribed picture element number over a picture data of a prescribed number in the data is extracted switchingly by the memory control means 2 and the two-dimension extrapolation predication discrete sine transform means 3 applies compression in the time base direction. Thus, the timewise correlation is utilized as to a consecutive data to attain highly efficient coding.

Description

[Detailed description of the invention]

(Industrial Application Field) The present invention relates to a three-dimensional image conversion and encoding device for converting and encoding images and compressing image data.

[Conventional technology]

As a method of performing highly efficient encoding using the correlation between pixels of an image, there is a method using extrapolation predictive discrete sine transformation, which the present inventors have already proposed in Japanese Patent Application No. 62-240455.

[Problem to be solved by the invention]

However, since this method uses the correlation between pixels in one piece of image data, it is inefficient when converting continuous image data because it does not use the temporal correlation of the image data. The present invention was invented in view of the above problems, and
The purpose is to provide a three-dimensional image transformation encoding device that can perform more efficient encoding by exploiting temporal correlation for continuous image data. Means for Solving the Problem] The present invention provides a memory that can store a plurality of consecutive image data, and an access order for the image data stored in this memory, either by a predetermined number of pixels in the horizontal or vertical directions of one screen, or by horizontally or vertically. A memory control double stage that can switch between a predetermined number of pixels in the vertical direction and a temporal direction, and a two-dimensional extrapolation prediction discrete sine conversion means that encodes image data by two-dimensional extrapolation prediction-discrete sine conversion. It is equipped with

[Effect]

In the apparatus of the present invention, the memory control means selectively extracts the prongs of predetermined lateral lines in the horizontal and vertical directions from the image data of each school, and performs predictive conversion encoding using the F stage of two-dimensional extrapolation predictive discrete sine conversion. After M-products of this predictive conversion encoded data are stored in memory, Glocks of a predetermined number of pixels covering a predetermined number of image data among the data are switched and extracted by the memory control means, and time is extracted by the two-dimensional extrapolation predictive discrete sine conversion means. It performs compression in the axial direction.

【Example】

The present invention will be explained below with reference to circuit configurations of embodiments. Figure 1 shows the configuration of the Ichiten embodiment, in which image data is
A memory 1 that can store +1 image, a two-dimensional extrapolation predictive discrete sine conversion means 3 described later, and a v! When passing data from the multiplied image data to the two-dimensional extrapolation predictive discrete sine transformation means 3, blocks of NXN pixels in the horizontal and vertical directions within one screen of image data are passed, or blocks of NXN pixels in the horizontal or vertical direction of the image data are passed. It is comprised of a memory control means 2 that switches whether to pass N pixels in the time direction or not. The two-dimensional extrapolation prediction discrete sine transformation means 3 divides the image into blocks as shown in FIG. a two-dimensional extrapolation prediction means 3a that subtracts the extrapolation prediction signal from the original image signal in the block and outputs it as an extrapolation prediction error signal; a conversion means 3b that performs two-dimensional discrete sign conversion on this extrapolation prediction error No. 46; A quantization means 3c cuts off the change coefficient transformed by the transformation means 3b at a certain threshold value and further quantizes it by a certain quantization step [[], and inverse quantizes the output of this quantization means 3c and performs two-dimensional inverse discrete sine transformation. and an inverse transform means 3d which reproduces the boundary value with the next block by performing addition with the extrapolated prediction signal.Since the image is transformed and encoded after being predicted, it is particularly difficult to use in the conventional method. It has the feature that the block size can be smaller than the DCTlj formula. As shown in FIG. 1, N+1 pieces of continuous image data are stored in the memory 1. Next, the memory control means 2 stores the Wm image data in the memory 1 one by one as shown in FIG.
The image is divided into XN image blocks, manually inputted to the two-dimensional extrapolation predictive discrete sine transform means 3, and predictive transform encoded. Predictive transform encoded data is stored in memory at the location of the most likely pixel data. After completing the predictive conversion encoding of each of the N+1 pieces of image data independently in this way, as shown in Fig. 3, the blocks of N×N pixels covering N pieces of the predictive conversion encoded data of the image data are r) is selected by the memory control means 2 and inputted to the two-dimensional extrapolation prediction discrete sine transformation means 3, where it is compressed in the time axis direction. At this time, the fourth option can select the vertical direction and time direction of the image data as shown in I, or it can also select the horizontal direction and time direction of the image as shown in FIG. Here, the uppermost image data portion is used as a boundary value in the time axis direction. Note that 4 is a practical value for N. FIG. 6 shows the circuit configuration of this embodiment used for predictive conversion encoding of moving images, and as shown in FIG.
Prepare two at i b and use this memory 1u. The input/output of 1b is switched by switching means 4a, 4b. The first moving image is input into the Nth image data area of the memo 'Jla, the subsequent image data for N images is input into the memory 1b, and the Nth image data stored in the memory 1a is inputted into the Nth image data area of the memo 'Jla. Predictive conversion coding is performed using the image data as a boundary value in the two-dimensional extrapolation predictive discrete sine conversion hand F13, and while the predictive conversion coding on the memo 1jlb side is being executed, images from the first to the N-1th image are Data is manually input to the memory 1a. memory 1
After the predictive conversion encoding of the image data on the ]] side is input to the memory 1b,
Predictive conversion encoding on the memory 1a side is started using this Nth image data as a boundary value. Thereafter, by repeating this procedure, predictive conversion encoding of moving images becomes possible. At this time, it is essential that the predictive transformation coding of N images be completed within the time period corresponding to N-1 moving images. .

【Effect of the invention】

Since the present invention is configured as described above, compression can also be performed in the time axis direction by utilizing the correlation regarding the passage of time in image data, and the compression rate can be increased compared to two-dimensional extrapolation predictive discrete sine transformation. Can be done.

[Brief explanation of the drawing]

Fig. m1 is a block diagram of an embodiment of the present invention, Fig. 2 is a block diagram of a two-dimensional extrapolation prediction discrete sign conversion means, and Fig. 3 is a diagram showing a memory scanning procedure for two-dimensional extrapolation prediction of image data. ,
FIG. 4 is a diagram showing a memory scanning procedure when performing extrapolation prediction in the time axis direction, FIG. 5 is a diagram showing another scanning method of FIG. 4, and FIG. 6 is a diagram showing another implementation of the present invention. It is a block diagram of an example. 1 is a memory, 2 is a memory control means, and 3 is a two-dimensional extrapolation prediction discrete sine conversion means. Agent Patent Attorney Ishi [J] Ai Shichizu 3 zu 4 zu 1116 zu

Claims (1)

[Claims] (1) A memory that can store multiple consecutive images, and an access order for the image data stored in this memory, either by a predetermined number of pixels in the horizontal or vertical direction of one screen, or by a predetermined number of pixels in the horizontal or vertical direction and in the time direction. It is equipped with a memory control means that can switch the number of pixels in increments, and a two-dimensional extrapolation prediction discrete sine conversion means that encodes image data by two-dimensional extrapolation prediction-discrete sine conversion. Blocks with a predetermined number of pixels in the horizontal and vertical directions are selectively extracted from the image data of , predictive transform coding is performed using a two-dimensional extrapolation predictive discrete sine transform means, and the predictive transform encoded data is stored in memory, and then the data is A three-dimensional image transformation characterized in that blocks of a predetermined number of pixels from a predetermined number of image data are selectively extracted by a memory control means and compressed in the time axis direction by a two-dimensional extrapolation predictive discrete sine transformation means. Encoding device.