Visual Dictionary Coding Reduces Data Size
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image and video coding methods are inefficient due to large data sizes even when using image features for compression, as they still require encoding and decoding of redundant pixel data.
Innovation Solution
A new coding and decoding method that establishes a visual dictionary of frequently occurring visual objects, extracts features, matches them using a feature matching method, and entropy codes the index and geometric parameters of matched visual words instead of the entire object, reducing data size and ensuring security by requiring the visual dictionary for decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image features are extracted and compressed instead of image pixels, then coding efficiency is improved, but the size of data content remains very large
Solution Approach 1:
The image is segmented into multiple visual objects, each of which is independently matched with visual words from a dictionary. This segmentation allows the system to process and encode only the essential semantic components rather than the entire image data, thereby reducing data content size while maintaining coding efficiency.
Solution Approach 2:
The invention extracts only the essential features of visual objects and represents them using indices to visual words from a pre-built dictionary, rather than encoding the complete image features. This extraction approach significantly reduces the data content size by eliminating redundant information while preserving the essential visual semantics.
2Quantity of substance
If the code stream is compressed to reduce data size, then storage and transmission efficiency is improved, but security vulnerability increases due to potential code stream capture
Solution Approach 1:
The visual dictionary serves as an intermediary component that is required for decoding but is not transmitted in the code stream. The code stream contains only indices and project parameters that are meaningless without the visual dictionary, creating a security mechanism where captured code streams cannot be decoded without access to the separate visual dictionary resource.
Solution Approach 2:
Instead of copying and transmitting the actual visual object data, the system uses indices as references to the visual dictionary. These indices are compact representations that point to the original visual words stored separately in the dictionary, reducing data size while enabling secure reconstruction when the dictionary is available.
3Productivity
If visual dictionary matching is used to represent objects, then coding efficiency is improved, but device complexity increases due to dictionary management
Solution Approach 1:
The visual dictionary is built in advance during an offline training phase, where visual objects are collected, processed, and organized into a structured dictionary with assigned indices. This preliminary action removes the complexity of dictionary creation from the runtime coding process, allowing the actual coding to efficiently query and use the pre-organized dictionary without managing its construction.
Solution Approach 2:
The visual dictionary serves multiple functions: it provides a reference for matching visual objects during coding, stores the actual visual representations for decoding, and acts as a security mechanism. This multi-functionality reduces the need for separate systems for each task, thereby managing complexity while improving overall coding efficiency.
Data Source
AI summary
A coding and decoding method for images or videos is provided by embodiments of the present invention to improve coding and decoding efficiency. The method includes: establishing a visual dictionary, wherein, the visual dictionary includes one or more visual words; extracting features from a specific object in an image; determining whether there is a visual word in the visual dictionary matching the specific object by using a feature matching method; obtaining the index of the visual word matched and a geometric relationship between the specific object and the visual word matched, wherein, the geometric relationship is represented by a project parameter; entropy coding the index of the visual word matched and the project parameter instead of entropy coding the specific object.