Image Stabilization for Temporal Noise Reduction in Image Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing apparatuses face limitations in capacity due to high memory usage, which prevents the accommodation of additional processing modules and effectively reduces image quality due to inadequate noise reduction capabilities.
Innovation Solution
An image processing apparatus and method that incorporates an image stabilization module to generate global motion vectors for temporal noise reduction, combined with spatial noise reduction, allowing for efficient processing and improved image quality without exceeding DRAM capacity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional image processing apparatus uses 2D spatial noise reduction only, then device complexity is low, but image quality is insufficient due to inadequate noise reduction
Solution Approach 1:
The patent divides noise reduction into two independent modules: temporal noise reduction module and spatial noise reduction module. The temporal module processes noise across multiple frames using motion vectors, while the spatial module processes noise within each frame. This segmentation allows each module to specialize in specific noise types without requiring the other to become overly complex, thereby improving image quality while maintaining manageable device complexity.
Solution Approach 2:
The image stabilization module serves dual purposes: it generates motion vectors for temporal noise reduction and performs traditional image stabilization. This multi-functionality reduces the need for separate dedicated modules, improving noise reduction capability without proportionally increasing device complexity.
2Manufacturing precision
If image processing apparatus adds more processing modules for noise reduction, then image quality improves, but DRAM capacity is exceeded
Solution Approach 1:
The patent performs motion estimation and generates motion vectors in advance during the image stabilization process, before temporal noise reduction is needed. This preliminary action allows the temporal noise reduction module to receive pre-computed motion vectors, eliminating the need for separate motion estimation computations and reducing overall memory requirements despite adding noise reduction functionality.
Solution Approach 2:
The patent uses the Y channel images from the image separation module as input for both image stabilization and temporal noise reduction, avoiding the need to store and process duplicate full-color images. This selective copying of only the necessary Y channel data reduces memory bandwidth and capacity requirements while maintaining effective noise reduction performance.
3Adaptability or versatility
If conventional apparatus stores Bayer pattern images in DRAM for processing, then processing can be performed, but DRAM capacity is occupied limiting accommodation of other modules
Solution Approach 1:
The patent extracts only the Y channel (luminance) information from the full-color YCbCr images for use in image stabilization and temporal noise reduction, rather than storing and processing complete RGB or Bayer images. This extraction of essential information reduces the memory footprint significantly, freeing up DRAM capacity to accommodate additional processing modules while maintaining the necessary processing capabilities.
Data Source
AI summary
An image processing apparatus and a processing method thereof are provided. The image processing apparatus includes an image capturing module, an image separation module, an image stabilization module, a temporal noise reduction module, and a spatial noise reduction module. The image capturing module captures a plurality of Bayer pattern images. The image separation module decreases the Bayer pattern images in size and transforms them into a plurality of YCbCr format images. The image stabilization module receives Y channel images of the YCbCr format images and the Bayer pattern images to perform motion estimation, to produce a plurality of global motion vectors (GMVs). The temporal noise reduction module performs temporal blending process on the Bayer pattern images according to the GMVs, to produce first noise reduction images. The spatial noise reduction module performs 2-dimensional spatial noise reduction on the first noise reduction images to produce second noise reduction images.