Electrical computers and digital data processing systems: input/ – Input/output data processing – Direct memory accessing
Reexamination Certificate
2001-01-11
2003-08-12
Shin, Christopher B. (Department: 2182)
Electrical computers and digital data processing systems: input/
Input/output data processing
Direct memory accessing
C710S022000, C711S005000, C345S537000
Reexamination Certificate
active
06606673
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a direct memory access transfer apparatus in which a data transfer between memories are directly performed through no central processing unit to process a large amount of picture data (or pixel data) relating to motion search or picture characteristic calculation performed in a picture processing system.
2. Description of Related Art
FIG. 10
is a block diagram showing the configuration of a picture processing system having a conventional direct memory access transfer apparatus. In
FIG. 10
,
100
indicates a memory of a data sender.
101
indicates a memory of a data receiver.
102
indicates a central processing unit (CPU).
103
indicates a conventional direct memory access (DMA) transfer apparatus.
104
indicates an address bus.
105
indicates a data bus.
Next, an operation of a picture processing system is described.
The case where picture data stored in the sender's memory
100
is sent to the receiver's memory
101
is, for example, described. A start address of the sender's memory
100
, a start address of the receiver's memory
101
and the amount of data to be transferred is sent from the CPU
102
to the conventional DMA transfer apparatus
103
, and an operation of the conventional DMA transfer apparatus
103
is started. In this operation of the conventional DMA transfer apparatus
103
, a readout address of the sender's memory
100
is output to an address bus
104
, and a readout signal is output from the conventional DMA transfer apparatus
103
to the sender's memory
100
. Thereafter, pixel data corresponding to the readout address in the sender's memory
100
is output from the sender's memory
100
to the data bus
105
.
The pixel data output to the data bus
105
is temporarily stored in a buffer of the conventional DMA transfer apparatus
103
. Thereafter, a write address of the receiver's memory
101
is output from the conventional DMA transfer apparatus
103
to the address bus
104
, and a write signal is output from the conventional DMA transfer apparatus
103
to the receiver's memory
101
. Thereafter, the pixel data stored in the buffer of the conventional DMA transfer apparatus
103
is output to the data bus
105
, and the pixel data output to the data bus
105
is stored in an area of the receiver's memory
101
indicated by the write address. Therefore, a data transfer between the sender's memory
100
and the receiver's memory
101
can be performed in the picture processing system according to the above-described processing.
For example, the case where a motion search processing in a picture coding is performed in the picture processing system is considered. In this processing, for a rectangular area of 16*16 pixels, all integral-numbered pixels placed from the −16-th relative pixel position to the +15-th relative pixel position are searched in the lateral and longitudinal directions. In this case, when a picture screen size having 352 (16*22) pixels in the lateral direction and 288 (16*18) pixels in the longitudinal direction is set at a frame rate of 30 Hz, the number of additions of differential absolute values between pixels required for every second is expressed according to a following equation.
(frame rate)*(the number of rectangular areas in one screen)*(the number of pixels in one rectangular area)*(the number of search operations)=30*(22*18)*16*16*(15+16+1)
2
≈3*10
9
To reduce a large number of addition operations expressed by the above equation, a following approximation calculation is generally known. In this approximation calculation, the pixels of each rectangular area are, for example, sub-sampled at the sample ratio of 2:1 according to a quincunx method to halve the number of additions of differential absolute values.
FIG. 11
is a conceptual view showing a sub-sampling performed at the sample ratio of 2:1 according to the quincunx method. As shown in
FIG. 11
, a group of pieces of pixel data of pixels arranged at even-numbered positions of one lateral line and a group of pieces of pixel data of pixels arranged at odd-numbered positions of another lateral line are alternately sampled for every lateral line.
This approximation calculation is not limited to the motion search. That is, in cases where a dispersed value of pixels of each rectangular area is calculated as a parameter indicating a characteristic of the rectangular area, an approximation calculation can be performed by sampling pieces of pixel data of the rectangular area at the sample ratio of 2:1 according to the quincunx method.
However, because the conventional DMA transfer apparatus has the above configuration, the conventional DMA transfer apparatus has no circuit mechanism in which pieces of pixel data are sub-sampled to reduce the number of calculations. Therefore, in cases where pieces of pixel data are, for example, sub-sampled at the sample ratio of 2:1 shown in
FIG. 11
, it is required to perform the sub-sampling of the pixel data according to a software processing performed in the CPU
102
. In this case, there is a problem that it takes a lot of time to perform the sub-sampling of the pixel data according to the software processing.
Also, in cases where the data transfer operation is performed for every pixel (for example, for every 8 bits) and an address producing method is adapted to the data transfer operation, a data sub-sampling and the transfer of sub-sampled data can be performed in the conventional DMA transfer apparatus. However, it is required to read pixel data from the sender's memory
100
for every pixel. Therefore, there is another problem that it takes a lot of time to transfer a large amount of pixel data.
SUMMARY OF THE INVENTION
An object of the present invention is to provide, with due consideration to the drawbacks of the conventional DMA transfer apparatus, a direct memory access transfer apparatus in which pieces of pixel data read out from a memory of a data sender are transferred to a memory of a data receiver at a high speed on condition that the accessing of the pieces of pixel data to each of the memories is possible at one time.
The object is achieved by the provision of a direct memory access transfer apparatus, in which pieces of pixel data are transferred between a sender's memory, to which the accessing of pieces of pixel data is possible at one time, a receiver's memory to which the accessing of pieces of pixel data is possible at one time, comprising a group of transfer parameter registers for storing a plurality of parameters required to transfer address data of both the sender's memory and the receiver's memory, a sender-memory control unit for producing a plurality of two-dimensional readout addresses of the sender's memory according to the parameters stored in the group of transfer parameter registers and reading out pieces of pixel data corresponding to pixels arranged on a plurality of lateral lines of a rectangular area from the sender's memory according to the two-dimensional readout addresses of the sender's memory, a data transforming unit for performing a sub-sampling for the pieces of pixel data of each lateral line read out from the sender's memory by the sender-memory control unit at a sample ratio of n:1 (n is an integral number higher than 1) and outputting pieces of sub-sampled pixel data corresponding to each lateral line, a receiver-memory control unit for producing a plurality of two-dimensional write addresses of the receiver's memory according to the parameters stored in the group of transfer parameter registers and write the pieces of sub-sampled pixel data of the lateral lines produced by the data transforming unit in the two-dimensional write addresses of the receiver's memory, and a timing control unit for controlling a processing timing of the sender-memory control unit, a processing timing of the data transforming unit and a processing timing of the re
Kamemaru Toshihisa
Ohira Hideo
Suzuki Hirokazu
Birch & Stewart Kolasch & Birch, LLP
Mitsubishi Denki & Kabushiki Kaisha
Shin Christopher B.
LandOfFree
Direct memory access transfer apparatus does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Direct memory access transfer apparatus, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Direct memory access transfer apparatus will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3125589