Automatic Image Trimming with Face and Composition Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic trimming methods for images often exclude backgrounds that are considered important by photographers, even if they are as significant as the subjects, leading to unwanted cropping of images.
Innovation Solution
An automatic trimming method and apparatus that detects facial images and judges the composition and position of the image to set a trimming frame that includes both the person and the background, using face detection, vertical direction judgment, composition discrimination, and facial-position judgment to determine the appropriate trimming frame based on the image's orientation and the photographer's intent.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If trimming is performed by detecting only facial images, then the automatic trimming process is simple and fast, but the background that the photographer considers important is excluded from the trimmed image
Solution Approach 1:
The patent segments the image analysis into multiple independent detection tasks: face detection, landmark detection, and vertical direction judgment. Each detection module operates independently to identify different elements (faces, landmarks, orientation), and their results are integrated to determine the final trimming frame. This segmentation allows comprehensive information capture while maintaining processing efficiency.
Solution Approach 2:
The trimming determination mechanism serves multiple functions: it considers face positions, landmark positions, and vertical direction judgments simultaneously to determine the appropriate trimming frame. This multi-functional approach ensures that both subject (face) and background (landmark) information are preserved while adapting to different composition orientations (portrait/landscape mode).
2Device complexity
If the trimming frame is set based only on face position, then the processing complexity is low, but the photographer's intent regarding background inclusion is not reflected
Solution Approach 1:
The patent implements dynamic trimming frame determination that adapts based on detected image characteristics. The system dynamically adjusts the trimming strategy according to the vertical direction judgment and the relative positions of faces and landmarks. This dynamic adaptation allows the system to reflect photographer intent in different scenarios without requiring complex manual configuration.
Solution Approach 2:
The system performs self-service by automatically judging the vertical direction and determining the appropriate trimming frame based on detected landmarks and face positions. The algorithm autonomously identifies whether the image is in portrait or landscape mode and adjusts the trimming parameters accordingly, eliminating the need for manual photographer input while preserving their intended composition.
3Speed
If automatic trimming is performed without considering image composition orientation, then the processing is faster, but the trimmed image may not maintain the intended portrait or landscape composition
Solution Approach 1:
The patent performs preliminary judgment of the vertical direction and image composition orientation (portrait or landscape mode) before determining the final trimming frame. By pre-establishing the orientation context, the system can quickly select the appropriate trimming strategy without complex real-time calculations, thus maintaining both processing speed and composition integrity.
Data Source
AI summary
Trimming is automatically performed on an image by inferring a person and a background in which the photographer of the image has interest. Facial images FP1 and FP2 are detected in a whole image P, and the vertical direction of the whole image P is judged based the facial images FP1 and FP2. Further, processing is performed using the information about the vertical direction to discriminate whether the composition of the whole image P is portrait-mode composition or landscape-mode composition. If it is judged that the facial image FP1 is positioned on the lower side of the whole image, and if a composition discrimination unit discriminates that the composition of the whole image P is landscape-mode composition, trimming is automatically performed using a trimming frame TG10 that has portrait-mode composition.