Image Processing Apparatus Pipeline Delay Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image decoding technologies face delays due to variations in size and shape of prediction units (PUs) and transform units (TUs), leading to inefficiencies in pipeline processing such as inverse frequency transform, motion compensation, and reconstruction processing.
Innovation Solution
An image processing apparatus that performs image processing for predictive images based on a predetermined transform order of transform units, allowing for efficient processing and reducing delays by aligning the processing order with the transform order, even when the prediction order differs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image processing is performed based on prediction unit order, then prediction processing can be completed, but processing delays occur due to size and shape variations of transform units
Solution Approach 1:
The image processing is divided into separate processing stages: prediction processing based on prediction unit order, and transform processing based on transform unit order. This segmentation allows each stage to operate independently with its own optimized processing order, eliminating delays caused by size and shape variations of transform units while ensuring prediction processing is completed correctly.
Solution Approach 2:
Prediction processing is performed in advance based on prediction unit order before transform processing begins. By completing prediction processing preliminarily and storing the results, the subsequent transform processing can proceed without waiting for prediction to finish, reducing overall processing delay.
2Productivity
If transform processing is performed before prediction processing, then pipeline efficiency can be improved, but processing order conflicts arise due to different unit sizes and shapes
Solution Approach 1:
The processing pipeline is segmented into independent prediction and transform stages, each with its own processing order rules. Prediction processing follows prediction unit order while transform processing follows transform unit order, eliminating conflicts between different processing orders and simplifying coordination complexity.
Solution Approach 2:
An intermediate storage buffer is introduced between prediction processing and transform processing. This intermediary allows prediction results to be stored and made available for transform processing without requiring synchronized processing orders, enabling independent optimization of each stage while maintaining data flow integrity.
3Loss of time
If processing order is changed to match transform unit order, then processing delays are reduced, but prediction processing must be reorganized
Solution Approach 1:
The processing organization is segmented into two independent ordering systems: prediction unit order for prediction processing and transform unit order for transform processing. This segmentation allows each processing type to maintain its natural order without forcing reorganization, reducing complexity while still enabling parallel and pipelined processing to reduce delays.
Solution Approach 2:
Prediction processing is performed in advance according to prediction unit order and results are stored in an intermediate buffer. This preliminary action allows transform processing to proceed according to transform unit order without needing to wait for or coordinate with prediction processing, reducing delays without complicating the processing organization.
Data Source
AI summary
An image processing apparatus for decoding an image coded in coding processes including (i) prediction on at least one prediction unit (PU) and (ii) frequency transform on transform units (TUs) included in a region including the at least one PU includes a processing unit configured to perform image processing for a predictive image of the at least one PU, depending on a transform order predetermined for the TUs.