Light Field Encoding Using Haar Wavelet Transform for Data Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for encoding and decoding light field data from plenoptic cameras are inefficient, leading to increased data storage and processing loads, as they either generate refocus images or depth maps in advance or decode data for each user request, failing to quickly adapt to various processing demands.
Innovation Solution
An image encoding and decoding technique that uses Haar wavelet transform to encode light field data by averaging pixel values and calculating difference values, allowing for efficient storage and decoding of only the necessary data, reducing redundancy and processing load by generating encoded data that can be quickly adapted to user requests for refocus, stereo vision, and depth estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If refocus images or depth maps are generated in advance from light field data, then processing speed for user requests is improved, but the amount of held data increases
Solution Approach 1:
The patent applies preliminary action by performing Haar wavelet transform and generating average images and difference images in advance during the encoding phase. These pre-processed components are stored in a compressed form that can be quickly combined later to generate refocus images or depth maps without storing the complete light field data set, thus achieving fast processing while maintaining data efficiency
Solution Approach 2:
The patent segments the light field data into multiple viewpoint images, then further segments each viewpoint image into average images and difference images through Haar wavelet transform. This segmentation allows selective decoding and processing of only the necessary components for each specific user request, reducing the amount of data that needs to be held and processed
2Quantity of substance
If light field data is encoded and decoded for each processing request, then the amount of held data is reduced, but the processing load increases
Solution Approach 1:
The patent performs preliminary Haar wavelet transform and organizes the data into average images and difference images during encoding. This pre-processing creates a structured format where subsequent decoding operations only require selective combination of pre-computed components rather than full light field decoding, significantly reducing processing load for each user request
Solution Approach 2:
The patent implements a dynamic processing approach where the decoding and combination of average images and difference images is adapted based on the specific user request (refocus, depth estimation, or stereo vision). This dynamic selection of processing paths optimizes computational efficiency for each operation type while maintaining data efficiency
3Quantity of substance
If conventional encoding methods (JPEG for one viewpoint, LZ for others) are used, then data storage is achieved, but the efficiency for various processing requests is insufficient
Solution Approach 1:
The patent creates a universal encoded data structure using Haar wavelet transform that serves multiple functions: refocus image generation, depth estimation, and stereo vision processing. The average images and difference images can be selectively combined to support all these processing types from a single encoded data set, providing adaptability and versatility that conventional single-purpose encoding methods cannot achieve
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This invention enables more efficiently storing and managing light field data. To accomplish this, when encoding N (N>1) viewpoint images captured by a light field imaging unit, an encoding apparatus calculates the average value of the values of N corresponding pixels corresponding to each other in the N viewpoint images and generates an image having the average value as the value of a pixel and focused at a predetermined focal position. The encoding apparatus generates (N - 1) types of difference values used to derive the values of the corresponding pixels in the N viewpoint images from the value of the pixel of the generated image focused at the predetermined focal position. The encoding apparatus encodes the generated image and difference information formed from the (N - 1) types of difference values.