google/guetzliArchived
Guetzli
Guetzli is a lossy image compression tool and perceptual JPEG encoder. It converts PNG or JPEG inputs into high-density JPEG files, reducing file size by removing data that the human eye cannot easily detect.
The tool utilizes human vision models to optimize the balance between file size and visual fidelity. It employs perceptual quality metrics and psychovisual similarity estimation to maintain high visual quality while maximizing compression density.
The project includes a visual difference analyzer capable of generating spatial difference heatmaps and calculating scalar similarity scores. These utilities allow for the comparison of compressed images against original reference sources to identify and quantify visual artifacts.
Features
- Image Compression Tools - Provides a high-density lossy compression tool that removes imperceptible data to reduce file size.
- Image Quality Assessment - Measures the perceived difference between original and compressed images to ensure visual quality.
- JPEG Conversion Utilities - Encodes PNG or JPEG inputs into high-density JPEG files to reduce storage size.
- High-Density Image Compression - Creates high-quality sequential JPEG images with increased compression density to minimize file sizes.
- Bitrate Optimization Algorithms - Repeatedly refines encoding parameters to maximize compression density while maintaining a perceptual quality threshold.
- Human Vision Models - Utilizes human vision models to prioritize the removal of image details invisible to the human eye.
- Perceptual Encoders - Creates high-density JPEG images that maintain visual quality through the use of human perception metrics.
- Perceptual Image Encoding - Compresses images into JPEG format using human vision models to minimize file size.
- Quantization Table Optimizations - Implements an exhaustive search of quantization tables to find the optimal balance between file size and quality.
- Image Optimization Tools - Utilizes perceptual quality metrics to reduce JPEG file size while maintaining high visual fidelity.
- Image Difference Heatmaps - Produces visual maps in PNM format to highlight specific spatial locations where two images differ most.
- Image Analysis - Provides tools for analyzing visual differences and generating heatmaps to identify compression artifacts.
- Perceptual Similarity Scoring - Calculates a single numerical value to quantify the overall perceived difference between an original and compressed image.
- Psychovisual Similarity Estimation - Calculates a scalar score and spatial map to measure perceived differences based on human visual perception.
- Reference-Based Quality Comparisons - Evaluates lossy compression results by comparing the output directly against the original uncompressed reference image.
- Spatial Difference Mapping - Generates a pixel-by-pixel map showing where visual artifacts occur to guide the compression process.
- Visual Difference Analyzers - Produces spatial maps to highlight the specific areas where two images differ most.
- Terminal Utilities - Perceptual JPEG encoder for file optimization.
