Cloud Image Compression Using Canonical Reference Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The rapid increase in image storage costs due to the vast number of photos hosted in cloud platforms poses a significant challenge, necessitating efficient image compression methods that balance compression rate and visual quality while minimizing storage needs.
Innovation Solution
The method involves identifying a canonical image set from existing cloud images, computing image representations, and using visually similar reference images from this set to compress new images using video compression techniques like H.265, leveraging binarized GIST descriptors for efficient k-nearest neighbors searches to maximize bit savings and maintain image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional image compression methods are used, then storage space is reduced, but visual quality deteriorates and compression rates are insufficient for large-scale cloud photo collections
Solution Approach 1:
The system pre-computes and stores image representations (features) for all images in the cloud collection before compression is needed. When a new image arrives, the system performs k-nearest neighbors search using these pre-computed representations to quickly identify visually similar reference images, enabling fast compression without sacrificing quality
Solution Approach 2:
Instead of storing redundant pixel data multiple times, the system creates a canonical image set where visually similar images are represented by copying their feature representations. The compression process reconstructs images by combining references from the canonical set, eliminating redundant storage while preserving visual quality through intelligent reference selection
2Quantity of substance
If cloud storage capacity increases to accommodate growing photo collections, then storage costs increase significantly
Solution Approach 1:
The system merges multiple visually similar images into a single canonical representation in the canonical image set. By combining redundant information from multiple sources into one standardized form, the system reduces the total storage capacity needed while maintaining the ability to reconstruct original images when needed
Solution Approach 2:
The system discards redundant pixel data from images that have visually similar content already present in the canonical set. When storage retrieval is needed, the system recovers the original image quality by reconstructing it from references in the canonical set, achieving space efficiency without permanent quality loss
3Quantity of substance
If image compression uses more reference images, then compression rate improves, but encoding and decoding time increases
Solution Approach 1:
The system pre-computes image representations and organizes them in the canonical image set before compression operations. This preliminary organization enables efficient k-nearest neighbors search using algorithms like Locality Sensitive Hashing, allowing the system to quickly identify reference images without performing computationally expensive operations during actual compression
Solution Approach 2:
The system uses a limited number of k nearest neighbors (partial action) from the canonical set rather than considering all possible reference images. This selective approach achieves sufficient compression quality with a manageable subset of references, avoiding the time penalty of processing excessive reference images while maintaining effective compression rates
Data Source
AI summary
Methods, systems, and computer readable media for performing image compression are disclosed. According to one exemplary method, the method includes identifying a canonical image set from a plurality of images uploaded to or existing on a cloud computing and/or a storage environment. The method also includes computing an image representation for each image in the canonical image set. The method further includes receiving a first image. The method also includes identifying, using the image representations for the canonical image set, one or more reference images that are visually similar to the first image. The method further includes compressing the first image using the one or more reference images.