Panoramic Image Compression via Non-Uniform Sampling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Panoramic images with ultra-high resolutions pose challenges in storage, transmission, and decoding due to large data sizes, and existing compression methods either compromise image quality or are inefficient in storage and transmission.

Innovation Solution

A method and device for panoramic image compression using non-uniform sampling, where a second spherical model is generated within a first spherical model based on a main view image, allowing for higher sampling rates for critical viewing angles and lower rates for less critical areas, resulting in a smaller compressed image.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If uniform compression is applied to the entire panoramic image, then the overall file size is reduced, but the image quality in the main viewing angle deteriorates

Engineering Contradiction:
Improvefile sizeVSAvoidimage quality
Core Design Contradiction:
Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent applies different compression rates to different regions of the panoramic image. Specifically, the main viewing angle region (corresponding to the second spherical model) is compressed at a lower rate to preserve quality, while other regions are compressed at higher rates to reduce file size. This is achieved by establishing different mapping relationships between the first spherical model (full panoramic image) and the second spherical model (main viewing angle), enabling region-specific compression strategies.

Inventive Principle:
Principle #3Local quality

2Quantity of substance

If high compression rate is used, then the file size is reduced, but the image definition deteriorates

Engineering Contradiction:
Improvedata sizeVSAvoidimage definition
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent segments the panoramic image into two distinct regions: the main viewing angle region (second spherical model) and other regions (first spherical model minus second spherical model). Each segment is then processed with appropriate compression rates. The main viewing angle region uses lower compression to maintain definition, while other regions use higher compression to reduce data size, thus resolving the contradiction between compression rate and image definition.

Inventive Principle:
Principle #1Segmentation

3Manufacturing precision

If the entire panoramic image is processed at full resolution, then the image quality is maintained, but the storage and transmission efficiency deteriorates

Engineering Contradiction:
Improveimage qualityVSAvoidstorage and transmission efficiency
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent changes the resolution parameter selectively across different regions of the panoramic image. The main viewing angle region is processed at a resolution that maintains quality while reducing data size through the spherical model transformation. Other regions are processed at lower resolutions. This parameter change strategy improves storage and transmission efficiency without significantly impacting perceived image quality in the main viewing angle.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3471052B1Panoramic image compression method and apparatus
Publication Date: 2021.06.09 BEIJING QIYI CENTURY SCI & TECH CO LTD
  • EP3471052B1 patent drawingFigure 1~2
  • EP3471052B1 patent drawingFigure 3~7
  • EP3471052B1 patent drawingFigure 8~9

AI summary

Embodiments of the present application provide a panoramic image compression method and device. The method comprises: obtaining a first spherical model formed by a first panoramic image to be compressed; generating a second spherical model in the first spherical model according to a main view image of an user; establishing a first mapping relationship between plane 2D rectangular coordinates in a second panoramic image and plane 2D rectangular coordinates in the first panoramic image; and sampling, from the first panoramic image, pixels corresponding to plane 2D rectangular coordinates in the second panoramic image according to the first mapping relationship to constitute the second panoramic image containing the pixels, so as to realize the compression of the first panoramic image. The method realizes the non-uniform compression of a first panoramic image, specifically, for images at a main viewing angle of users, a higher sampling rate is set so as to guarantee the definitions of images at a main viewing angle of users, and for the other images, a lower sampling rate is utilized so as to compress the first panoramic image into a smaller second panoramic image and facilitate the storage, transmission and decoding of the panoramic image.