Hybrid CNN Background Detection for Portrait Replacement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional chroma key technology is limited in replacing backgrounds in portrait photographs, as it requires specific saturated colors and struggles with distinguishing subjects from backgrounds when the subject's clothing is similar in color, and it does not effectively handle backgrounds or floors with patterns or varied colors.
Innovation Solution
A system and method using machine learning to detect and classify photographic scenes, including backgrounds and floors, by training a hybrid deep learning model with convolutional neural networks to extract color and texture features, allowing for the automatic identification and replacement of photographic scenes without relying on saturated colors or patterns, enabling the use of a wide range of colors and designs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If chroma key technology is used to replace backgrounds, then background replacement can be achieved with specific saturated colors, but it cannot handle backgrounds with patterns or varied colors and fails when subject clothing is similar in color to the background
Solution Approach 1:
The patent transforms the background detection approach from relying on specific color parameters (chroma key's saturated blue/green) to using multiple image processing parameters including color histograms, edge detection, and pattern recognition. This allows the system to detect backgrounds with varied colors and patterns while maintaining subject-background distinction through multi-parameter analysis rather than single-color dependency
Solution Approach 2:
The patent combines multiple detection methods (color analysis, edge detection, pattern recognition) into a composite detection system. This composite approach integrates different processing techniques to achieve both adaptability to various backgrounds and precision in subject-background distinction, overcoming the limitations of single-method chroma key technology
2Productivity
If conventional chroma key technology is used, then the process is simple and fast, but it is limited to saturated colors and cannot accurately detect backgrounds with patterns or varied colors
Solution Approach 1:
The patent performs preliminary analysis of the background characteristics (color distribution, patterns, edges) before executing the replacement operation. By pre-processing the image to identify background regions and characteristics, the system maintains efficient processing speed while enabling detection of diverse backgrounds that chroma key cannot handle
3Ease of operation
If chroma key technology requires specific saturated colors, then the detection process is straightforward, but it fails when the subject's clothing is similar in color to the selected background color
Solution Approach 1:
The patent moves from single-dimensional color-based detection (chroma key's single color parameter) to multi-dimensional detection incorporating color histograms, spatial edges, texture patterns, and contextual information. This dimensional expansion allows the system to maintain operational simplicity while achieving reliable subject-background distinction even when colors are similar
Data Source
AI summary
A method of identifying a background type in a photograph includes extracting a background image from a photograph, feeding the background image into a first convolution neural network to obtain a first decision, extracting color features in the background image, transforming the color features into a two-dimensional color feature matrix, feeding the two-dimensional color feature matrix into a second convolution neural network to obtain a second decision by the one or more computer processors, extracting texture features in the background image, transforming the texture features into a two-dimensional texture feature matrix image by the one or more computer processors, feeding the two-dimensional texture feature matrix into a third convolution neural network to obtain a third decision, computing a hybrid decision based on the first decision, the second decision, and the third decision, and identifying a background type in the background image based on the hybrid decision.