Multi-modal Tone-mapping Using Gaussian Mixture Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image mapping techniques, such as histogram equalization and single sigmoid functions, fail to effectively reduce high bit-depth images to lower bit-depth while preserving detail and avoiding artifacts, particularly in medical imaging where fine details and dynamic range are crucial.
Innovation Solution
A multi-modal image mapping system using a Gaussian mixture model to compute statistics and generate sigmoid functions for transforming high bit-depth images to lower bit-depth, allowing for efficient and accurate preservation of image details and dynamic range.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If histogram equalization is used to compress dynamic range, then bit-depth reduction is achieved, but fine details are lost and artifacts are introduced
Solution Approach 1:
The patent segments the histogram into multiple modes representing different tissue types or intensity ranges. Each mode is processed separately with its own mapping function, allowing fine details within each segment to be preserved while achieving overall compression. This is done by identifying peaks in the histogram and creating separate transformation functions for each mode rather than applying a single global transformation.
Solution Approach 2:
The patent applies different mapping characteristics to different regions of the histogram corresponding to different tissue types. Each local region (mode) receives a customized transformation that preserves its specific details and characteristics, rather than applying a uniform transformation across the entire dynamic range. This local optimization prevents the loss of fine details in specific tissue regions.
2Loss of information
If local histogram equalization is used to preserve details, then fine details are maintained, but computational complexity increases significantly
Solution Approach 1:
The patent divides the image processing task into segments based on intensity modes identified in the histogram. Instead of processing every pixel with a complex local window operation, the image is segmented into regions corresponding to different modes, and each segment is transformed using a pre-computed mapping function. This segmentation approach preserves details while reducing computational complexity compared to full local histogram equalization.
Solution Approach 2:
The patent performs preliminary analysis of the histogram to identify modes and pre-compute the mapping functions before applying the transformation to the actual image data. This preliminary action allows the complex transformation logic to be prepared in advance, reducing the computational burden during the actual image processing stage and enabling real-time or near-real-time performance.
3Productivity
If a single sigmoid function is used for tone-mapping, then computational speed is improved, but detail preservation across all luminance levels is insufficient
Solution Approach 1:
The patent combines multiple sigmoid functions, each optimized for different intensity ranges or tissue types, into a composite mapping function. This composite approach maintains the computational efficiency of sigmoid functions while achieving superior detail preservation across the full dynamic range. The composite function is constructed by weighting and combining individual sigmoid functions based on the identified histogram modes.
Data Source
AI summary
A system for multi-modal mapping of images is described. Embodiments are described where the image mapping system is used for visualizing high dynamic range images such as medical images, satellite images, high dynamic range photographs and the like and also for compressing such images. In examples, high bit-depth images are tone-mapped for display on equipment of lower bit-depth without loss of detail. In embodiments, the image mapping system computes statistics describing an input image and fits a multi-modal model to those statistics efficiently. In embodiments, the multi-modal model is a Gaussian mixture model and a plurality of sigmoid functions corresponding to the multi-modal model are obtained. In an embodiment the sigmoid functions are added to form a tone-mapping function which is used to transform a high bit-depth image such as 16 or 12 bits per pixel to a low bit-depth image such as 8 bits per pixel.