Intra-Prediction Mode Selection for Flat Areas in Moving Image Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing moving image encoding devices face challenges in improving image quality, particularly in areas with flat pixel values like windowboxing, due to differences between input and local decoded images, leading to suboptimal prediction mode selection and image degradation.
Innovation Solution
The solution involves a moving image encoding device that uses flat area determination parameters, such as activity and sum of absolute differences, to calculate offset values for prediction mode determination, allowing for improved selection of intra-prediction modes and reducing image degradation in motionless areas with flat pixel values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If intra-prediction mode determination uses an input image, then encoding speed is improved through parallel processing, but image quality deteriorates due to inability to select optimal intra-prediction mode in flat pixel value areas
Solution Approach 1:
The patent calculates flat area determination parameters (activity and sum of absolute differences) in advance for all blocks before mode determination. This preliminary calculation allows the system to identify flat pixel value areas ahead of time, so that when intra-prediction mode determination is performed using the input image for speed, the optimal mode can still be selected by referencing these pre-calculated parameters. This resolves the contradiction by preparing necessary information beforehand without delaying the speed-critical path.
2Object-affected harmful factors
If DC mode is forcibly changed in areas with large difference between input image and local decoded image, then image degradation is suppressed, but advantageous effect of improving image quality is insufficient
Solution Approach 1:
The patent applies different prediction mode selection strategies based on local block characteristics. For blocks identified as having flat pixel values (through activity and sum of absolute differences calculations), the system selects from multiple intra-prediction modes including planar and angular modes. For non-flat blocks, the DC mode or other appropriate modes are used. This localized adaptation to block characteristics resolves the contradiction by providing image quality improvement where needed (flat areas) while maintaining efficiency elsewhere.
3Manufacturing precision
If intra-prediction mode is determined using local decoded image, then optimal intra-prediction mode can be selected, but encoding speed is impaired due to inability to perform parallel processing
Solution Approach 1:
The patent pre-calculates flat area determination parameters (activity and sum of absolute differences) for all blocks before the actual mode determination step. This allows the mode determination to be performed using the input image (enabling parallel processing for speed) while still having the necessary information to select optimal modes for flat areas. The preliminary calculation of characteristics decouples the speed-critical path from the quality-optimization path.
Solution Approach 2:
The flat area determination parameters (activity and sum of absolute differences) serve as intermediary data that bridges the gap between using input image for speed and local decoded image for quality. These pre-calculated parameters provide the necessary information about block characteristics without requiring the actual local decoded image data during the speed-critical mode determination step, thus acting as a mediator that enables both speed and quality optimization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A moving image encoding device encoding a moving image by using a computation unit subtracting a prediction image from an input image for each pixel block, comprising: a prediction mode determination unit selecting an intra-prediction mode, from the plural intra prediction modes, that generates a prediction image using pixel block adjacent to coding target block, wherein the prediction mode determination unit calculates a flat area determination parameter evaluating a flatness of each pixel block of the input image, calculates for a coding target block an intra prediction mode offset value for each intra prediction mode based on the flat area determination parameter of an adjacent block used for the intra prediction of the coding target block of the input image, calculates an evaluation value evaluating coding cost of the coding target block for each intra prediction mode, and selects the intra prediction modes having a smallest intra evaluation value which is obtained by adding respective intra prediction mode offset values and the corresponding evaluation values.