The entire disclosure of Japanese Patent Application No. 2005-373209 including the specification, claims, drawings and abstract is incorporated herein by reference.
1. Field of the Invention
The present invention relates to a decoding apparatus that obtains encoded image data and executes decoding processing on the input image data.
2. Description of the Related Art
The H.264 video compression standard can achieve very high data compression in a video encoding/decoding system. Similar to other standards known as MPEG2 and MPEG4, H.264 is based on an encoding technique using motion compensation or inverse conversion or spatial region ←→ frequency region conversion/quantization processing. An H.264 compatible system executes decoding processing for each block with a deblocking filter that can remove block distortion from decoded data.
An H.264 decoding system uses pre-filter data for intra-prediction (prediction based on data in a frame) and uses post-filter data for inter-prediction (prediction based on data of plural frames).
If a frame memory is “temporarily” used for intra-prediction, because the filter processing cannot be started until the intra-prediction is completed, the filter processing is postponed. To deal with this problem, a decoder may include a decoded image memory storing intra-prediction data in addition to a frame memory, so that the filter processing can be executed without any delay.
A decoded image memory storing apparatus 16 receives the decoded image data from the adder 14 and causes a decoded image memory 18 to store the decoded image data. A filter 20 receives the decoded image data from the decoded image memory 18 and applies filter processing to the decoded image data. A frame memory storing apparatus 22 receives filter processed data (hereinafter, referred to as “Post-filter data”) from the filter 20. The frame memory storing apparatus 22 causes a frame memory 24 to store the post-filter data.
A decoded image memory reading apparatus 26 can read non-filtered data (hereinafter, referred to as “pre-filter data”) from the decoded image memory 18. An intra-prediction section 28 receives the pre-filter data from the decoded image memory reading apparatus 26. The intra-prediction section 28 produces an intra-prediction signal based on the pre-filter data.
A frame memory reading apparatus 30 can read post-filter data of plural frames from the frame memory 24. An inter-prediction section 32 receives the post-filter data of plural frames from the frame memory reading apparatus 30. The inter-prediction section 32 produces an inter-prediction signal based on the post-filter data of plural frames.
A selection switch 34 receives the intra-prediction signal from the intra-prediction section 28 and the inter-prediction signal from the inter-prediction section 32. The selection switch 34 can selectively supply the intra-prediction signal or the inter-prediction signal to the adder 14, so that the adder 14 can produce decoded image data based on the inter-prediction signal or the intra-prediction signal.
The decoding apparatus of
For example, an example of the H.264 decoding apparatus is described by Yi Hu et al. in “Development of HDTV compatible H.264 video decoder IP core” (Design Wave Magazine, pp. 105-111, August 2004).
The above-described decoding technique can simultaneously and smoothly execute both the filter processing and the intra-prediction. However, a decoding apparatus based on this technique must include a large-scale memory for storing one frame of decoded image data. Furthermore, the decoding apparatus requires a circuit for reading decoded image data from the memory and a circuit for storing decoded image data into the memory. Thus, the circuit scale of the decoding apparatus grows disadvantageously larger.
The present invention can constantly store, in a frame memory, pre-filter data required for intra-prediction. The present invention can realize effective data storage without requiring a frame memory storing pre-filter data used for intra-prediction.
The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate an embodiment of the invention and, together with the description, serve to explain the principles of the invention, in which:
Embodiments of the present invention will be described with reference to the drawings.
An entropy decoding section 10 inputs a bit stream of encoded data and produces an entropy decoded signal. An inverse quantization/inverse conversion section 12 applies inverse quantization and inverse conversion to the entropy decoded signal. An adder 14 receives an output of the inverse quantization/inverse conversion section 12. The adder 14 also receives inter-prediction data or intra-prediction data. Thus, the adder 14 can produce decoded image data reflecting inter-prediction or intra-prediction.
A frame memory storing apparatus 22 receives the decoded image data (i.e., pre-filter data) from the adder 14. The frame memory storing apparatus 22 writes the pre-filter data into a frame memory 24. A frame memory reading apparatus 30 can read the pre-filter data from the frame memory 24. A filter 20 receives the pre-filter data read from the frame memory reading apparatus 30. The filter 20 applies deblocking filter processing to the pre-filter data. The frame memory storing apparatus 22 writes the filtered data (i.e., post-filter data) into the frame memory 24.
The frame memory reading apparatus 30 supplies predetermined post-filter data of plural frames to an inter-prediction section 32. The inter-prediction section 32 produces an inter-prediction signal based on the received post-filter data. Furthermore, the frame memory reading apparatus 30 reads predetermined pre-filter data from the frame memory 24 and supplies the read pre-filter data to an intra-prediction section 28. The intra-prediction section 28 produces an intra-prediction signal based on the received pre-filter data.
A selection switch 34 receives the intra-prediction signal from the intra-prediction section 28 and the inter-prediction signal from the inter-prediction section 32. The selection switch 34 can selectively supply the intra-prediction signal or the inter-prediction signal to the adder 14, so that the adder 14 can produce decoded image data based on the inter-prediction signal or the intra-prediction signal. A control circuit 36 can control the actions of the frame memory storing apparatus 22, the frame memory reading apparatus 30, and the selection switch 34.
First, as shown in
Namely, the decoding apparatus of the present embodiment overwrites the post-filter data on the pre-filter data in an offset manner. More specifically, when the post-filter data are written into the target storage region of the frame memory 24, a peripheral region of the pre-filter data block can remain without being overwritten by the post-filter data. The peripheral region remaining unchanged corresponds to at least one row and one column (two rows and two columns according to the example shown in
When the above-described processing is repeated, the frame memory 24 can store pre-filter data in a peripheral region including a row extending along lower sides of consecutively processed blocks and a column extending along the right side of the last-processed block, as shown in
The intra-prediction section 28 executes prediction based on 1-column pre-filter data extending along the left side and 1-row pre-filter data extending along the upper side of the intra-prediction data as shown in
A plurality of prediction modes may be employed. In the present embodiment, the intra-prediction section 28 uses 1-row pre-filter data extending along the lower side of the immediate upper-layer block and 1-column pre-filter data extending along the right side of the neighboring block (i.e., last processed block).
As shown in
The frame memory 24 stores post-filter data of all processed blocks. Therefore, the frame memory reading apparatus 30 can read necessary data from the frame memory 24 and supply the read post-filter data to the inter-prediction section 32.
As described above, the system according to present embodiment offsets a storage region of pre-filter data relative to a target storage region in the frame memory 24 and overwrites post-filter data on the target storage region. Accordingly, the frame memory 24 can hold both the post-filter data required for the inter-prediction and the pre-filter data required for the intra-prediction. The invention as embodied in the present embodiment can designate necessary pre-filter data and hold the specified pre-filter data only. Accordingly, with the present embodiment, the amount of pre-filter data which must be stored for the prediction can be reduced. Furthermore, the processing required in the present embodiment, which involves changing the writing position, is relatively simple to execute.
In the present embodiment, the data temporary holding circuit 40 can store pre-filter data of a target block to an offset region as illustrated in the above-mentioned
Furthermore, a device other than the filter 20 can be provided to extract pre-filter data required for the intra-prediction. The extracted pre-filter data can be written to a predetermined portion of the frame memory 24. The post-filter data, when obtained, can be written into the frame memory 24. Thus, the frame memory 24 can store the data as shown in
The intra-prediction performed according to the present invention is not limited to a specific type. The pre-filter data used for the prediction is not limited to 1-column/1-row data. Therefore, if another type of data is used, the method for storing (leaving) necessary pre-filter data in the frame memory 24 can be appropriately selected.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2005-373209 | Dec 2005 | JP | national |