Image Region of Interest Detection via Discrete Cosine Transform

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for finding regions of interest in images and photos are not effective for still images, as they are based on motion analysis and do not work well for managing and visualizing collections of images on various displays.

Innovation Solution

An algorithm that analyzes sub-blocks of an image using the discrete cosine transform for information content and compressibility, grouping low compressibility sub-blocks into regions of interest through a morphological technique, applicable to arbitrary images and photos, with a center-weighted variation for improved results in photo applications.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If motion analysis-based algorithms are used to find regions of interest, then video segmentation can be achieved, but the method does not work for still images

Engineering Contradiction:

Improveapplicability to different image typesVSAvoideffectiveness for still images

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent changes the fundamental parameter used for ROI detection from motion-based metrics to information content-based metrics using discrete cosine transform. This allows the algorithm to work effectively on still images by measuring compressibility and information density rather than motion, thereby achieving versatility across different image types while maintaining reliability

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent creates a universal algorithm that can handle both video and still images by using a general-purpose information content measurement approach. The discrete cosine transform-based method serves multiple functions: it works for video frames, still photos, and various display sizes, making the system adaptable without sacrificing effectiveness

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If sub-blocks are analyzed using discrete cosine transform for information content, then regions of interest can be identified in arbitrary images, but the computational complexity increases

Engineering Contradiction:

Improvegenerality to arbitrary imagesVSAvoidalgorithm complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent divides the image into sub-blocks and analyzes each sub-block independently using discrete cosine transform. This segmentation approach allows the complex DCT operation to be applied to smaller, manageable units rather than the entire image, reducing overall computational complexity while maintaining the ability to identify ROIs in arbitrary images

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies DCT analysis selectively to sub-blocks rather than processing the entire image uniformly. By focusing computational effort on dividing the image into manageable sub-blocks and analyzing only those regions, the algorithm achieves generality for arbitrary images while keeping complexity manageable through partial application of the transform

Inventive Principle:

Principle #16Partial or excessive action

3Ease of operation

If collections of images are managed and visualized on small displays, then mobile viewing is enabled, but image detail and quality are compromised

Engineering Contradiction:

Improveviewing on mobile devicesVSAvoidimage quality

Core Design Contradiction:

Ease of operationVSManufacturing precision

Solution Approach 1:

The patent extracts only the essential regions of interest from each image using information content analysis, rather than attempting to display the entire image. By identifying and displaying only the most informative sub-blocks grouped into ROIs, the system enables mobile viewing on small displays while preserving image quality and detail in the extracted regions

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments images into sub-blocks and identifies key regions for display, allowing selective presentation of image content. This segmentation enables efficient use of small display space while maintaining quality by focusing on the most important visual information rather than compressing or downsizing the entire image

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS7724959B2Determining regions of interest in photographs and images

Publication Date: 2010.05.25 FUJIFILM BUSINESS INNOVATION CORP

AI summary

An algorithm for finding regions of interest (ROI) in images and photos based on an information driven approach in which sub-blocks of an image are analyzed for information content or compressibility based on the discrete cosine transform. The sub-blocks of low compressibility are grouped into ROIs using a morphological technique. Unlike other algorithms that are geared for highly specific types of ROI (e.g. face detection), the method of the present invention is generally applicable to arbitrary images and photos. A center-weighted variation of the algorithm can produce better results for certain photo applications. The algorithm can be used with several other image applications, including Stained-Glass collages and Pan-and-Scan presentations.