Image Classifier for Face Detection in Low-Quality Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face detection methods struggle with low-quality images, particularly those affected by light conditions and motion artifacts, leading to reduced recognition rates and incorrect detections, such as mistaking backgrounds for faces or failing to detect faces in low-light or blurred images.
Innovation Solution
An image processing apparatus utilizing convolutional neural networks (CNNs) to classify images as high-quality or low-quality, identify the clearest images among low-quality ones, and perform face detection on the clearest images, while also classifying and evaluating blurred face images to improve detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing face detection methods use high-quality images for training, then detection accuracy is improved, but detection rate for low-quality images significantly decreases
Solution Approach 1:
The patent applies preliminary action by performing image quality classification and evaluation before face detection. The system first classifies images as high-quality or low-quality using a CNN classifier, then evaluates low-quality images to identify the clearest ones. This preliminary sorting enables the face detection algorithm to process images in an optimal sequence, improving both accuracy and detection rate by ensuring that the most detectable images are processed first.
2Reliability
If preprocessing is performed to eliminate interferences in low-quality images, then detection rate is improved, but device complexity increases due to multiple preprocessing types required
Solution Approach 1:
The patent implements universality by creating a unified image processing framework that handles multiple types of low-quality images (backlight, low-light, blurred, motion artifacts) through a single integrated system. The CNN-based image quality classifier and evaluator provide a universal preprocessing approach that automatically adapts to different image quality issues without requiring separate preprocessing pipelines for each image type, thereby reducing device complexity while maintaining improved detection rates.
Data Source
AI summary
An image processing apparatus, includes an image classifier configured to determine whether an input image is a low-quality image or a high-quality image; and an image evaluator configured to determine a first predetermined number of clearest images from a plurality of low-quality images determined by the image classifier.