Block-Based Digital Refocusing System for Portable Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital refocusing methods face challenges in achieving real-time, low-computation, and high-quality refocusing, particularly on portable devices, due to high computational demands and issues like discontinuity in foregrounds and distortion with large aperture changes.
Innovation Solution
A block-based digital refocusing system and method that divides images into block data, performs refocusing computations on regional data, and combines results to achieve efficient refocusing, using view interpolation and alpha blending to reduce blocking effects and optimize computation based on depth maps and disparity maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If ray-tracing method is used for digital refocusing, then refocusing quality is improved, but computation complexity increases making real-time refocusing impossible
Solution Approach 1:
The image is divided into multiple depth layers based on depth information, with each layer processed independently through focused image guidance. This segmentation allows complex refocusing to be broken into manageable per-layer operations, reducing overall computational complexity while maintaining refocusing quality.
Solution Approach 2:
Depth information is extracted from the input image to create depth layers, separating the refocusing computation from processing the entire image at once. By extracting and processing only relevant depth layers with focused image guidance, the method reduces computation complexity compared to full-ray-tracing while preserving refocusing quality.
2Manufacturing precision
If light field method with view interpolation is used, then real-photo quality is achieved, but computation complexity remains too large for portable devices
Solution Approach 1:
The image processing is segmented into depth-based layers, where each layer is processed with focused image guidance rather than applying complex light field computations to the entire image. This reduces computation complexity for portable devices while maintaining real-photo quality through selective processing of relevant depth regions.
3Manufacturing precision
If adaptive filter is used to generate depth information, then blur kernel size is correlated with depth value, but discontinuity phenomenon appears in foreground when background is in focus
Solution Approach 1:
Focused image guidance acts as an intermediary between depth information and the final refocused image. By using the focused image to guide the processing of each depth layer, the method maintains foreground continuity and avoids discontinuity phenomena that occur with adaptive filtering alone, while still utilizing depth-correlated blur kernel sizes.
4Device complexity
If layered depth map is used for hierarchical processing, then computation is reduced, but distorted phenomenon appears in picture with large aperture changes
Solution Approach 1:
Different processing strategies are applied to different depth layers based on their local characteristics. Focused image guidance is applied selectively to each depth layer, allowing the method to reduce computation complexity through hierarchical processing while minimizing picture distortion by adapting processing quality to local depth regions.
Data Source
AI summary
A block-based digital refocusing method includes a capturing step, a dividing step, a selecting step, a refocusing step and a combining step. The capturing step is for capturing at least one picture datum. The dividing step is for dividing the picture datum into a plurality of block data. The selecting step is for defining a regional datum according to each of the block data. The refocusing step is for conducting a refocusing computation to obtain a refocused block datum according to each of the regional data. The combining step is for combining each of the refocused block data based on each of the regional data to form a refocused picture datum.