Image Path Bundling for Resizing Overhead Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image resizing methods using 'seam carving' face challenges in ensuring accurate reinsertion of paths at the same position as deleted paths, leading to encoding overhead, loss of relative positioning information, or weak resizing, particularly in video sequences where temporal aspects are critical.
Innovation Solution
A method that groups deleted paths into bundles based on specific criteria, such as equi-distributed areas and content-based selection of lines, and uses an energy map combining saliency and gradient maps to determine optimal path deletion and reinsertion, ensuring efficient bandwidth usage while preserving image quality and object positioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If path position information is transmitted for each deleted path, then image reconstruction quality is improved, but encoding overhead and bandwidth consumption increase significantly
Solution Approach 1:
The patent merges multiple deleted paths that are close to each other into a single representative path. Instead of transmitting position information for each individual path, only the representative path's position is encoded and transmitted. This significantly reduces the quantity of path information that needs to be transmitted while still enabling accurate reconstruction of all deleted paths in the region.
Solution Approach 2:
Instead of transmitting detailed position information for every deleted path (which increases bandwidth), the patent inverts the approach by transmitting minimal information about representative paths and using interpolation algorithms to reconstruct the positions of all other paths. This inverts the relationship between transmission data量和reconstruction accuracy.
2Productivity
If paths are removed to achieve compression, then bandwidth efficiency is improved, but relative positioning information of objects is lost
Solution Approach 1:
The patent performs preliminary grouping of paths based on their spatial relationships and characteristics before compression. By identifying and grouping paths that are close to each other and belong to the same structural region, the system preserves the relative positioning information of objects while achieving compression. The grouping criteria ensure that paths representing object boundaries and important features are maintained.
Solution Approach 2:
The patent applies different treatment to different regions of the image based on local characteristics. In regions with important object boundaries or features, paths are preserved with higher fidelity. In uniform or less important regions, paths can be more aggressively compressed or represented by fewer representative paths. This local quality approach maintains relative positioning information where it matters most.
3Device complexity
If straight lines only are removed for resizing, then device complexity is reduced, but manufacturing precision and visual quality deteriorate due to diffusion function requirements
Solution Approach 1:
The patent transitions from static straight-line path removal to dynamic curved path removal. The paths can now follow the contours and boundaries of objects in the image, adapting to the local geometry. This dynamic approach allows paths to bend and curve around objects, preserving their shapes and relative positions more accurately while maintaining reasonable algorithmic complexity through greedy optimization.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for encoding a digital image or sequence of digital images, comprising the steps of: a) deleting (30) paths between two opposite edges of an image using an algorithm for resizing by deleting pixels, in order to define a reduced image; and b) modeling (34) the deleted paths for transmission with the reduced image, with a view to interpolating an expanded image. The step (34) of modeling the deleted paths comprises: b1) grouping (35) the deleted paths into bundles of paths according to at least one predetermined criterion; and b2) generating (36) a description of the bundles of deleted paths.