Image Segmentation Using JPEG Artifact Resilient HSV Layering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing background removal methods for still images in embedded applications face challenges due to JPEG compression artifacts, which affect segmentation accuracy and computational complexity, making them unsuitable for real-time applications.
Innovation Solution
A model that sorts pixels by similar brightness values and models the background as a convex polygon, allowing for efficient segmentation by determining which pixels belong to the object of interest based on their position relative to the polygon, even in images subjected to lossy compression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard background removal methods are used for still images in embedded applications, then segmentation performance can be improved, but computational complexity increases making them unsuitable for real-time applications
Solution Approach 1:
The patent segments the image processing task by first identifying background pixels and then separating foreground objects based on color similarity to background. This divides the complex segmentation problem into manageable steps: background identification, color space transformation, and pixel classification, reducing overall computational complexity while maintaining accuracy
Solution Approach 2:
The patent transforms pixel data from RGB color space to HSV (Hue, Saturation, Value) color space to improve segmentation accuracy. This parameter transformation allows for better color-based discrimination while the method processes only necessary pixels (background and similar-colored pixels), reducing computational load compared to processing all pixels
2Productivity
If JPEG compression is applied to images from embedded cameras, then image transmission and storage efficiency is improved, but segmentation accuracy deteriorates due to compression artifacts
Solution Approach 1:
The patent acknowledges that JPEG compression artifacts affect segmentation accuracy but works within this constraint by using color-based segmentation in HSV space. The method accepts compressed images as input and uses color similarity metrics that are relatively robust to compression artifacts, converting the limitation into an acceptable trade-off for real-time processing
3Device complexity
If color-based segmentation is performed in RGB color space, then processing simplicity is maintained, but segmentation accuracy deteriorates compared to HSV color space
Solution Approach 1:
The patent explicitly transforms pixel values from RGB to HSV color space to improve color segmentation accuracy. The HSV color space separates color information (hue) from brightness (value), allowing for more accurate color-based background removal. This transformation is applied only to pixels that need processing, balancing accuracy improvement with computational efficiency
Data Source
AI summary
A process to isolate objects of interest in an image uses a model which takes into account the artifacts stemming from the JPEG compression. The pixels are sorted accordingly to layers which regroup pixels of similar value, and the background is modeled by a polygon from an additional picture. Segmentation based on the knowledge of the background and the layers is then performed. This method provides good results for segmentation, while being sufficiently fast to be implemented for real time application.